The Syntax of Givenness Ivona Kucerová

The Syntax of Givenness
by
Ivona Kučerová
Submitted to the Department of Linguistics and Philosophy
in partial fulfillment of the requirements for the degree of
PhD
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2007
c○ Massachusetts Institute of Technology 2007. All rights reserved.
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Linguistics and Philosophy
May 25,2007
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Danny Fox
Associate Professor of Linguistics
Thesis Supervisor
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sabine Iatridou
Professor of Linguistics
Thesis Supervisor
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alec Marantz
Professor of Linguistics
Thesis Supervisor
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
David Pesetsky
Professor of Linguistics
Thesis Supervisor
Accepted by. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Irene Heim
Professor of Linguistics, Chair of the Linguistics Program

Abstract
The Syntax of Givenness
by
Ivona Kučerová
Submitted to the Department of Linguistics and Philosophy
on May 25,2007, in partial fulfillment of the
requirements for the degree of
PhD
The goal of this thesis is to account for distributional patterns of given and new items in
Czech, especially their word order. The system proposed here has four basic components:
(i) syntax, (ii) economy, (iii) interpretation, and (iv) reference set computation. The approach
belongs to the family of interface driven approaches.
The syntactic part of the thesis introduces a free syntactic movement (G-movement).
The movement causes very local reordering of given elements with respect to new elements
in the structure. G-movement is licensed only if it creates a syntactic structure which leads
to a semantic interpretation that would not otherwise be available. The economy condition
interacts with the way givenness is interpreted. I introduce a recursive operator that adds
a presupposition to given elements. The distribution of the operator is regulated by the
Maximize presupposition maxim of Heim (1991). The reference set for purposes of this
evaluation is defined as the set of derivations that have the same numeration and the same
assertion.
Finally, I argue that the licensing semantic conditions on givenness in Czech are not
identical to the licensing conditions on deaccenting in English. The givenness licensing
conditions are stronger in that they require that for an element to be given it must not only
have a salient antecedent but also satisfy an existential presupposition.
Thesis Supervisor: Danny Fox
Title: Associate Professor of Linguistics
Thesis Supervisor: Sabine Iatridou
Title: Professor of Linguistics
Thesis Supervisor: Alec Marantz
Title: Professor of Linguistics
Thesis Supervisor: David Pesetsky
Title: Professor of Linguistics
2

Acknowledgments
Two things were difficult about writing my thesis: (i) choosing the topic (there were too
many things I would like to write a thesis about) and (ii) choosing an adviser. There were
several people I wanted to have as my adviser. In the end I decided to write my thesis on
four different topics and to have four advisers. The first decision turned out to be untenable
and I ended up dropping three of the topics. The second decision, on the other hand, paid
off; except for typographical issues while trying to typeset the cover page of my thesis I
have never regretted it.
Alec, Danny, David and Sabine (in alphabetical order) are more than wonderful scholars.
They have been my mentors and guides without ever telling me what to do. In a sense,
they are coauthors of many ideas in my thesis: most of the ideas took their shape while
sitting together for hours, playing with puzzles and trying different ways to attack them. I
have really enjoyed the eight months I have spent writing my thesis.
My stay at MIT was of course more than my dissertation. It is impossible to thank all the
people I have met here and whom I have learned from. Among my teachers a special place
belongs to Noam Chomsky, Mike Collins, Morris Halle, Irene Heim and Donca Steriade. I
want to thank them here for their time and interest in my work.
Sometimes it is hard to draw a clear line between a friend and a colleague. Many people
were both. Asaf Bachrach, Roni Katzir and Raj Singh became my ‘classmates’ and
provided me with intense underground education and great food. Brooke Cowan, Svea
Heinemann and Sarah Hulsey made sure I would know that life at MIT is more than linguistics.
Special thanks go to Béd’a Jelínek without whom I would have never applied to MIT.
He knows best that there is more I should thank him for.
I want to thank also my former teachers without whom I would have not even thought
about starting a PhD: Vlasta Dvořáková, Jindřich Rosenkranz, Jaroslava Janáčková, Michael
Špirit, Jiří Homoláč and †Alexander Stich (in chronological order).
There is no way I could thank my parents and Roni Katzir without trivializing their
place in my life.
3

Contents
1 To Be Given 6
1.1 G-movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2 Basic word order and Focus projection . . . . . . . . . . . . . . . . . . . . 17
1.3 Verb partitions and their semantic ambiguity . . . . . . . . . . . . . . . . . 22
1.4 The boundedness of G-movement . . . . . . . . . . . . . . . . . . . . . . 26
1.5 The distribution of pronouns: G-movement as a last Resort . . . . . . . . . 34
1.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2 G-movement 41
2.1 The target of G-movement . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.2 When in the derivation does G-movement apply? . . . . . . . . . . . . . . 45
2.3 The head movement restriction on G-movement . . . . . . . . . . . . . . . 56
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3 Complex Word Order Patterns 64
3.1 Deriving the verb partition . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2 Multiple G-movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4 Semantic Interpretation 86
4.1 Where we stand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.2 Marking givenness by an operator . . . . . . . . . . . . . . . . . . . . . . 93
4.2.1 G-movement domains are propositional domains . . . . . . . . . . 94
4.2.2 The G-operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.3 The evaluation component . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.4 The G-operator and coordination . . . . . . . . . . . . . . . . . . . . . . . 114
4.5 The interpretation of Givenness . . . . . . . . . . . . . . . . . . . . . . . . 120
4.6 Givenness versus deaccentuation . . . . . . . . . . . . . . . . . . . . . . . 123
4.7 A note on phonology and its relation to givenness . . . . . . . . . . . . . . 128
4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
A G-movement is A-movement 133
A.1 Condition A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
A.2 Condition C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
A.3 The Weak Cross-Over Effect . . . . . . . . . . . . . . . . . . . . . . . . . 139
4

A.4 A note on base generation . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
A.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5

Chapter 1
To Be Given
Consider the Czech sentences in (1). 1,2
(1) a. SVO: Chlapec našel lízátko.
boy.Nom found lollipop.Acc
b. OVS: Lízátko našel chlapec.
lollipop.Acc found boy.Nom
c. SOV: Chlapec lízátko NAŠEL.
boy.Nom lollipop.Acc found
d. OSV: LÍZÁtko chlapec našel.
lollipop.Acc boy.Nom found
The sentences in (1) describe a similar situation. In each of them the speaker asserts that
there was a time interval in the past such that an event of finding took place in that interval.
Furthermore, we learn that the event of finding had two participants (a finder and a findee)
and we also learn who these participants were (some boy and a lollipop). The sentences in
(1) nevertherless differ in their meaning and in the set of contexts in which they are felicitous.
To see this, let’s first concentrate on the first two orders, i.e., the SVO and the OVS
orders. Corresponding English translations of the Czech sentences are given in (2). 3
1 Capital letters here and throughout the text stand for a contrastively stressed syllable.
2 The four combinations given in (1) are the only combinations that can be used as declarative clauses. The
remaining permutations, given in (i) and (ii), are grammatical but only as questions. The following chapters
will deal with declarative clauses only.
(i)
(ii)
VSO: Našel chlapec lízátko?
found boy.Nom lollipop.Acc
‘Did the boy found the lollipop?’
VOS: Našel lízátko chlapec?
found lollipop.Acc boy.Nom
‘Was it the lollipop what the boy found?’
3 The hash sign # stands for an utterance that is not felicitous in the given context. In this particular case,
it stands for an infelicitous translation.
6

(2) a. SVO: Chlapec našel lízátko.
boy.Nom found lollipop.Acc
(i) ‘A boy found a lollipop.’
(ii) ‘The boy found a lollipop.’
(iii) ‘The boy found the lollipop.’
(iv) #‘A boy found the lollipop.’
b. OVS: Lízátko našel chlapec.
lollipop.Acc found boy.Nom
‘A boy found the lollipop.’
As we can see in (2-a), the SVO order in Czech is compatible with several different interpretations.
The very same Czech string can correspond (i) to a situation where neither a boy,
nor a lollipop are given, in the sense that there is no referent determined by the previous
context, i.e., the existence of the referent has not been asserted yet; (ii) to a situation where
only a boy has been previously determined by the context; or (iii) to a situation where both
the boy and the lollipop have a unique referent but they have not been introduced by the
previous context. Crucially, however, the SVO order is not felicitous in a situation in which
only the lollipop has been introduced by the previous context, (iv).
To achieve the missing interpretation, i.e., the interpretation in which only the object
has been determined by the previous context, the word order must be OVS, as in (2-b),
translated as ‘A boy found the lollipop.’.
What is the nature of the reordering? Notice that in order to capture the intuition about
meaning differences corresponding to different word orders, I used indefinite and definite
articles. Since Czech does not have any overt morphological marking of definiteness – with
the exception of demonstrative and deictic pronouns – we could understand the different
word orders as a strategy to achieve the same interpretation that English can achieve by
using overt determiners. This would be, however, a simplification. The object in the OVS
order does not need to correspond to a definite description. It is enough that it has been
introduced in the previous discourse, as in (3). 4
(3) a. We left some cookies and lollipops in the garden for the kids. Who found a
lollipop?
b. Lízátko našla Maruška a Janička.
lollipop.Acc found Maruška and Janička
‘Little Mary and little Jane found a lollipop.’
4 The reordering observed in (2-a) and (2-b) might remind the reader of the Mapping Hypothesis of Diesing
1992 or of a more general discussion of specificity as in Enç 1991; van Geenhoven 1998; Farkas 2002, among
many others. One might think that for a DP to become specific (whatever it means) such a DP must move to
(or it must be base-generated in) a certain syntactic position. As we will see shortly, not only referential, but
also predicational or propositional elements can be introduced in a discourse in the same way as the object in
(2-b). I do not know at this point whether there is any connection between specificity and the data discussed
here. In general, I will ignore possible relations between quantification and information structure here.
7

We will see in chapter 4, however, that the correlation with definite articles is not accidental.
I will argue that the purpose of both reordering for givenness and definiteness marking
is to maximize presupposition (Heim, 1991). While the two strategies are in principle different,
their realization may sometimes be identical.
I thus argue that the nature of the reordering found in (2-b) is indeed to mark that
something has already been introduced in the discourse. An interesting fact to observe
is that even though the SVO order has multiple interpretations, certain interpretations are
excluded and they can be achieved only by reordering. This suggests that the purpose of
the reordering is to achieve an interpretation that would not be available otherwise. 5
I will argue shortly that this ordering correlates with a requirement that given (old, presupposed)
elements linearly precede elements that are new (asserted, non-presupposed) in
the discourse. While in (2-a) this requirement can be satisfied within the SVO order, in
case of the (2-b) examples, reordering is needed.
The examples in (1-c) and (1-d), repeated below as (4-a) and (4-b), are rather different.
While in the examples in (2-a) and (2-b) at least one participant has not been introduced in
the discourse yet, the examples in (4-a) and (4-b) contain only previously introduced participants.
In other words, the utterances in (2-a) and (2-b) correspond to an assertion that
combines something already presupposed with something that has not been presupposed
yet. In contrast, the utterance in (1-c)–(1-d) operates on an already asserted proposition.
The meaning of the SOV order in (4-a) can be either verification of the true value of the
proposition, or correction of a part of the proposition by excluding other alternatives. 6
Similarly, the example in (4-b) corresponds to an utterance about already introduced participants
and it comments on their relation (contrastive topic or topic).
(4) a. SOV: The boy did find the lollipop. (He did not steal it.)
b. OSV: As for the lollipop, the boy found it (but as for the chocolate, he got it
from his mother).
The study presented here concerns mainly the type of examples given in (2). Before I
undertake a closer investigation of these cases I want to provide the reader with some further
observations that will help to clarify why it is useful to treat the examples in (2) as separate
from the examples in (4). We have already seen that the type of reordering witnessed in
(2) may teach us something about the way presupposed and non-presupposed elements are
marked in a language that does not have overt articles. Everything being equal, the English
translations of the Czech examples in (2) differ only in the use of articles. There are no
other obvious changes in syntax, prosody or morphology. The English translations of the
Czech examples in (4) are very different in this respect. Both have marked prosody (pitch
5 We will see in this chapter and especially in section 1.2 that SVO is the basic word order in Czech. I will
argue that a basic word order in general allows more flexibility in interpretation than a derived word order.
6 In literature on information structure, this type of construction is often referred to as verum focus, i.e.,
a construction where the speaker verifies or denies that an already presupposed proposition is true, or contrastive
focus or correction, in case the speaker asserts an exclusion of an alternative that might have been
introduced by the previous context.
8

accent on the auxiliary did in (4-a) and pitch accent on the lollipop in (4-b), both followed
by optional deaccenting). Furthermore, the example in (4-a) contains additional morphological
material (did) and the example in (4-b) may correspond in English to a cleft, i.e., to
a rather complex structure. I suggest that Czech is like English in that the processes that
lead to the reordering in Czech (2-b) are of a very different nature than the processes that
lead to the reorderings witnessed in Czech (4). If we want to understand these processes
we should study them separately. 7
There is another parallel between Czech and English worth mentioning. It concerns
differences between the prosodic properties of (2) and (4). As in English, the Czech intonation
of the examples in (2) is neutral. Furthermore, there are no relevant differences
between (1-a) and (1-b). Compare figure 1-1 and 1-2. In contrast, the intonation of (4) is
marked. In (4-a), the verb našel ‘found’ is distinct in its duration and intensity. As we can
see in figure 1-3, the duration of the bisyllabic verb is roughly doubled in comparison with
the pronunciation of the same word in other syntactic environments. In (4-b), the sentence
initial object lízátko ‘lollipop’ is pronounced with a higher tone and the sequence is followed
by deaccenting and an optional intonational break, as in figure 1-4.
500
400
300
200
Pitch (Hz)
100
0
chlapec nasel lizatko
0 2.83288
Time (s)
Figure 1-1: SVO order
7 Reordering processes, superficially similar to those seen in (2) and (4), are often referred to as scrambling.
The term originated in Ross 1967 as a term for stylistic reordering. In the more recent literature,
scrambling may refer to A-movement, A-bar movement or base generation (for example, Williams 1984;
Saito 1989; Webelhuth 1989; Grewendorf and Sternfeld 1990; Mahajan 1990; Lasnik and Saito 1992; Saito
1992; Corver and van Riemsdijk 1994; Miyagawa 1997; Bošković and Takahashi 1998; Bailyn 2001; Karimi
2003; Sabel and Saito 2005). I will avoid the term scrambling here in order to minimize possible confusion
that might come from various interpretations the term has been assigned in the literature.
9

Pitch (Hz)
500
400
300
200
100
0
lizatko nasel chlapec
0 3.32218
Time (s)
Figure 1-2: OVS order
Before we get to the actual proposal there is another empirical observation to be established.
Notice that in the examples in (2) the verb occupies the second position. In contrast,
in the examples in (4) the verb linearly follows the arguments. 8 As we will see shortly, two
factors turn out to be crucial for understanding Czech word order variations: the relative
position of arguments and the relative position of the verb with respect to the arguments.
In the rest of this chapter and in the following chapters I will closely investigate the
nature of the reordering witnessed in (2). I will argue that this type of reordering is derived
by short A-movement that is sometimes parasitic on verb head movement. I will call this
movement G-Movement. This movement happens, I will claim, only in order to derive a
semantic interpretation that would not be available otherwise. As such this kind of movement
is driven by interpretation requirements and is restricted by economy.
The goal of section 1.1 is to introduce the basics of a system that can account for the
Czech word order data observed in (2). I will present the proposal in stages.
The first version of the system, introduced in this section, is designed to account for
the order of the arguments and the position of the verb. Further refinements will be introduced
in the following chapters. In the next section, section 1.2, I will address the question
of how come multiple interpretations are available (only) for basic word orders, such as
SVO. In particular, I will argue that the ambiguity follows from G-movement being a last
resort operation. Section 1.3 will more closely investigate the fact that the verb intervenes
between the two arguments, in contrast to other types of reorderings, such as questions
8 To make the picture complete, in questions the verb precedes the arguments.
10

Pitch (Hz)
500
400
300
200
100
0
chlapec lizatko nasel
0 3.39374
Time (s)
Figure 1-3: SOV order
500
400
300
200
Pitch (Hz)
100
0
lizatko chlapec nasel
0 3.18576
Time (s)
Figure 1-4: OSV order
11

and contrastive foci readings. I will argue that even though G-movement is independently
restricted by syntax like any other type of movement, it differs from other types of movement
in that it is parasitic on head movement. The emergence of verb second will be seen
as a consequence of this restriction. In section 1.4 I will explore further consequences of
this restriction, namely, differences in locality restrictions on G-movement dependent on
head movement properties of the relevant head. In the last section of this chapter, 1.5, I
will provide independent evidence for the proposed machinery. I will argue based on the
distribution of pronouns that G-movement is a last resort operation. As such it takes place
only if the relevant interpretation would not otherwise be available.
1.1 G-movement
The idea that word order in a language such as Czech is sensitive to the discourse in the
sense that parts of a clause that are old (given, in the background etc.) linearly precede the
elements that are new in the discourse has been investigated in Czech linguistics for a long
time. 9
The questions that have not, to my knowledge, been successfully addressed yet are (i)
how exactly is the linear partition (i.e., given ≻ new) derived, and (ii) how is the word order
within the old and the new part determined.
Consider the example in (5). This sentence could be an answer to ‘What did Petr do
yesterday with his car?’ The || sign corresponds to the partition between given and new.
Now compare (5) and (6). The example in (6) shows the basic word order and as such it is
a suitable answer to the question ‘What happened?’. 10 As we can see schematized in (7),
the two word orders differ quite radically.
(5) Petr auto včera || řídil rychle.
Petr.Nom car.Acc yesterday drove fast
‘Yesterday Petr drove his car fast.’
←− S O Adv1 V Adv2
9 The relevance of the discourse for word order has already been observed by traditional grammarians (for
example, Gebauer 1900). In the pre-modern tradition, the observation was stated in terms of psychological
and structural subjects. To my knowledge, the intuition about the broader relevance of the discourse was
for the first time formalized in a more systematic way by Vilém Mathesius in Mathesius [1929] 1983 and
Mathesius 1939 (the first version of the paper was presented in Prague in 1908). In order to describe the
relation between word order and the discourse, Mathesius introduced the term aktuální členění. This term is
usually translated as topic focus articulation. This is not an exact translation though. The literal translation
would be something like structuring [of the sentence] dependent on the current context and referred to theme
versus rheme distinction. Authors that further developed the notion of word order dependence on the current
context include, for example, Daneš 1954, 1957, 1974; Novák 1959; ; Dokulil and Daneš 1958; Šmilauer
1960; Adamec 1962; Firbas 1964; Hausenblas 1964; Sgall 1967; Benešová 1968; Hajičová 1973, 1974; Sgall
et al. 1980, 1986; Hajičová et al. 1998, among many others. There is no way I can acknowledge in this study
all of the empirical observations, generalizations and linguistic insight that is present in the previous work on
aktuální členění in Czech.
10 I will provide diagnostics for determining basic word order in 1.2.
12

(6) Petr řídil včera rychle auto.
Petr.Nom drove yesterday fast car.Acc
‘Yesterday Petr drove fast his car.’
←− S V Adv1 Adv2 O
(7) Petr drove yesterday fast car −→ Petr car yesterday || drove fast
There are two facts to be noticed. (i) The given elements in (5) precede the new elements.
(ii) While the relative order of the new items in (5) is preserved relative to (6), the relative
order of the given items has changed. The changes are highlighted in (8) and (9).
(8) Word order of the given elements:
a. Basic: Petr drove yesterday fast car
b. Derived: Petr car yesterday || drove fast
(9) Word order of the new elements:
a. Basic: Petr drove yesterday fast car
b. Derived: Petr car yesterday || drove fast
As we will see throughout the coming sections and chapters, this is a typical pattern
that arises in other places as well: The relative order of given items with respect to the
basic word order may undergo various permutations, while the relative order of new items
usually stays unchanged. The only exception is in fact a new finite verb, as we will see
shortly. In all other cases, new items do not change their relative order at all. As we will
see later in this section, this observation is important and it will become crucial for the
way we are going to account for different Czech word order patterns. As we will see, reorderings
happen in a principled way and they can be directly predicted from the relevant
syntactic structure.
Another important fact is that the relative order of given items only sometimes changes.
As we can see in (8), the relative order of the subject Petr with respect to ‘car’ and ‘yesterday’
does not change. In contrast, the order of the object auto ‘car’ and the adverbial včera
‘yesterday’ is reversed. For example, in the same sentence without the adverb ‘yesterday’,
the relative word order of the given items would be the same, (10).
(10) Petr auto || řídil rychle.
Petr.Nom car.Acc drove fast
‘Petr drove his car fast.’
I propose that these word order facts suggest that only given elements move for information
structure purposes. This is a simplistic picture because new finite verbs sometimes move
as well but let’s stay with the simple generalization for the sake of building the proposal in
stages.
13

Notice that there is no optionality in the word order of given elements in a sentence
with a particular meaning. Thus, we need to have a restrictive syntactic system that would
account for the word order. I argue that given elements undergo a special kind of movement
that I will call G-movement. The rules governing G-movement are given in (11). Further
restrictions on G-movement are stated in (12).
(11) G-Movement [version 1]
G-movement must take place
a. iff α G is asymmetrically c-commanded by a non-G element,
b. unless the movement is independently blocked.
(12) Restrictions on G-movement:
G-movement is restricted as follows:
a. α G moves to the closest position X, such that no non-G element asymmetrically
c-commands α G .
b. If α is XP, then α moves to an XP position.
c. If α is a head, then α moves to an X 0 position.
(13) Closeness: (after Rizzi (1990))
X is the closest to Y only if there is no Z such that Z c-commands Y and does not
c-command X.
Following Reinhart 1997, 2006; Fox 1995, 2000, I argue that G-movement is a syntactic
operation that takes place only if it affects one or both of the interfaces. In particular, I
argue that G-movement must have semantic import. In other words, the grammar I argue
for is restricted by economy in that it allows only syntactic operations that lead to a
distinct semantic interpretation. Notice that if there is no non-G element asymmetrically
c-commanding α G the closest position that satisfies the requirement on G-movement is the
position of α G itself. Thus, if there is no structurally higher new element, α G does not
move.
The definition of G-movement implies that an element does not enter the computation
marked as given but it is only the result of the computation that the element is interpreted as
such. As we will see in 1.5, this property is crucially connected to the fact that G-movement
is a last resort operation.
Furthermore, (11) crucially relies on the notion of asymmetrical c-command (Kayne,
1994) 11 and it does not distinguish heads from phrases, in the sense that both heads and
11 The relevant definitions are given below:
(i) X asymmetrically c-commands Y iff X c-commands Y and Y does not c-command X. (Kayne, 1994,
p. 4, (2))
(ii)
X c-commands Y iff X and Y are categories and X excludes Y and every category that dominates X
dominates Y. (Kayne, 1994, p. 16, (3))
(iii) In the sense of Chomsky 1986, p. 9: X excludes Y if no segment of X dominates Y. (Kayne, 1994, p.
133, ftn.1)
14

phrases are required to undergo G-movement. 12 Consider the trees in (14).
Taking into account that Czech is an SVO language, there are two basic cases to consider.
Either (i) α G is a head and β (a non-G element) is a phrase, as in (14-a); or (ii) α G is
a phrase and β is a head, as in (14-b).
(14) a.
βP α G
b.
β
α G P
In case of (14-a), the definition of G-movement requires the head α G to move above
the phrase, resulting in (15). I leave open for now what exactly the landing site of such a
movement is. I will address this question in section 2.3.
(15)
α G
βP t αG
An example of such movement is found in unergatives. In Czech, in the basic word
order an unergative subject precedes an unergative verb, as in (16). If the verb is given and
the subject is new, the word order is reversed, as in (17). 13
(16) a. What happened?
b. Marie tancovala.
Marie danced
(17) a. Who danced?
b. Tancovala || Marie.
danced Marie
‘Marie danced.’
In case of (14-b), the definition of G-movement requires the phrase α G to move over the
head β, resulting into (18).
(18)
α G P β tαG
A simple case of such movement can be found with unaccusatives. In contrast to unergatives,
in the basic word order, an unaccusative subject follows an unaccusative verb, as in
(19). If the verb is new and the subject is given, the word order is reversed, as in (20).
12 The proposal predicts that if there were rightward movement, a given element might follow a new element.
Unfortunately, I am not aware of any case of rightward movement in Czech that would allow to test
this prediction.
13 I simplify the derivation here. In the full derivation, V moves to v, resulting into structure that requires
another instance of G-movement. I will go through derivations in more details in the coming chapters. For
now, I leave many details aside.
15

(19) a. What happened?
b. Přijel vlak.
arrived train
‘A train arrived.’
(20) a. What happened to the train?
b. Vlak || přijel.
train arrived
‘The train arrived.’
This is basically the story. 14 There are still many questions that need to be addressed –
and I will address them in the coming sections and chapters – but the core of the argument
is as simple as this: If a given element α G is asymmetrically c-commanded by a non-G
element, α G needs to move. If there is no offending c-command relation, G-movement
does not take place.
Before I approach developing the syntactic system in more detail, I want to shortly address
what I mean by given.
In the literature on information structure it is agreed, at least since Halliday (1967), that
an utterance may be divided into two parts, one of which is more established in the discourse
(common ground, context. . . ) than the other. While the terminology and the actual
approaches vary widely (see for example Kruijff-Korbayová and Steedman (2003) for a
recent overview), there is a strong intuition that the two parts are complementary to each
other. Thus, it is a reasonable assumption that a grammar might refer only to one of the
parts. The question then is which part is relevant.
The approach I take here is based on the idea that the part encoded in the grammar is
the given part (on given/new or given/focus scale). For arguments for this type of approach
see, for instance, Williams (1997); Schwarzschild (1999); Krifka (2001); Sauerland (2005);
Reinhart (2006).
As to the question of how exactly given is defined for now I will adopt Schwarzschild’s
definition of givenness, stated in (21). The definition is based on the observation that
for element α to be given, which in English roughly corresponds to being deaccented, α
must be entailed by the previous discourse and α must have a salient antecedent. Since
entailment refers to proposition, the definition ensures that entailment can be stated for any
α (not only propositional but also referential and predicational), (21-b).
(21) Definition of Given (modified from Schwarzschild 1999, p. 151, (25))
α is interpreted as Given in an utterance U iff there is a salient antecedent A such
that
a. α and A corefer; or
14 There is one crucial piece missing: dependence of G-movement on head movement. I will discuss this
property in section 1.3 and 2.3.
16

. A entails (α,U) where (α,U) is an utterance derived by replacing all maximal
constituents of U (except for α) by variables existentially quantified over.
Any other element is a non-G element.
The definition will be more than sufficient for the discussion of the syntax of givenness in
chapters 1–3. We will see, however, in chapter 4 that the proposed syntactic mechanism
covers only a subset of the relevant cases. For the rest of the Czech empirical pattern we
will need to combine the syntactic system with a semantic one. In chapter 4 we will also
see that the Schwarzschild definition is too weak for the Czech facts. But all this must wait.
Let’s concentrate on the syntax first.
1.2 Basic word order and Focus projection
This section shows that only basic word orders are compatible with multiple interpretations.
In order to account for this fact, I will argue that the lack of ambiguity for derived orders
follows from G-movement being a last resort operation. If a structure is such that there
is no new (non-G) element asymmetrically c-commanding given α G , no G-movement is
expected. The core idea that I will introduce in this section is that while a structure without
any instance of G-movement may be compatible with several interpretations, G-movement
always disambiguates.
So far I have not established how we define the basic word order. I propose, following
Veselovská (1995); Junghanns and Zybatow (1997); Lavine and Freidin (2002); Bailyn
(2003), among others, that the basic word order in Czech is the all-new word order. This
means that the basic word order is an order that can be used as a felicitous answer to the
question What happened?. 15 An example of such an order is given in (22). Notice also that
binding, which is sometimes used for determining basic word order, cannot be used as a
diagnostic for Czech. The reason is that Slavic languages have A-movement scrambling
that can change the basic word order and create new binding configurations. I will comment
on binding properties of G-movement in appendix A.
(22) Basic word order:
A: What happened at the party?
B 1 : [Marie dala Pavlovi facku] New .
Marie.Nom gave Pavel.Dat slap.Acc
B 2 : #Marie dala facku Pavlovi.
Marie.Nom gave slap.Acc Pavel.Dat
B 3 : #Pavlovi dala Marie facku.
Pavel.Dat gave Marie.Nom slap.Acc
B 4 : #Facku dala Marie Pavlovi.
slap.Acc gave Marie.Nom Pavel.Dat
‘Marie slapped Pavel. [literary: Marie gave Pavel a slap.]’
15 The cited authors address only Slavic languages, more precisely Czech, Ukrainian, and Russian.
17

It has also been observed that in basic word order various constituents may be understood
as new. The only condition for an SVO language like Czech is that a new constituent be
aligned to the right edge of a clause. We can test this property by using wh-questions
targeting different constituents, as seen in (23).
18

(23) What can be understood as new?
a. (i) What did Marie give to Pavel?
(ii) Marie dala Pavlovi [facku] New ←− slap
b. (i) What did Marie give to whom?
(ii) Marie dala [Pavlovi facku] New ←− Pavel a slap
c. (i) What did Marie do?
(ii) Marie [dala Pavlovi facku] New ←− gave Pavel a slap
d. (i) What happened?
(ii) [Marie dala Pavlovi facku] New ←− Marie gave Pavel a slap
In contrast, if an utterance contains any deviance from the basic order, then such an utterance
is infelicitous in an all-new context. It means that if any reordering takes place,
at least one constituent must be α G , i.e., introduced in the previous discourse. In other
words, any reordering limits the number of structural positions in which we can identify
the partition between given and new. For example, in a derived word order, as in (24), there
is only one felicitous interpretation of the information structure. More precisely, only the
rightmost constituent can be interpreted as new (non-G). In this particular case, it is the
indirect object Pavel. Thus, in (24) there is only one possible partition between given and
new, in contrast to (22) that is compatible with several partitions, as schematized in (25).
(24) Focus Projection within a derived word order:
a. Marie dala facku [Pavlovi] New ←− S V DO || IO t DO
b. #Marie dala [facku Pavlovi] New
c. #Marie [dala facku Pavlovi] New
d. #[Marie dala facku Pavlovi] New
(25) a. Marie dala facku || Pavlovi.
b. (||) Marie (||) dala (||) Pavlovi (||) facku.
What we can learn from the observed pattern is that in basic word order there is a relative
freedom in what parts of the utterance can be interpreted as new and what parts can be interpreted
as given. As I have already anticipated in the previous discussion of G-movement,
this pattern can be described in the following manner: whatever is interpreted as given cannot
be linearly preceded by anything interpreted as new.
Let’s now turn to the question of how exactly the multiple partition effect follows from
our system. To see that, we will look at a very simple case: a transitive clause that has no
modifiers, only a subject, a verb, and an object. Consider first the case where the subject is
the only given element. As we already know, the resulting word order is SVO, as seen in
(26) and (27). 16
(26) a. Subject-G verb Object ←−
b. #Object verb Subject-G
16 I use here examples with potom ‘then, afterward’ instead of a wh-question. The reason is that potom
creates a natural context where only the subject is presupposed/given. For reasons that are not clear to me, it
is difficult to obtain the same pragmatic effect with a wh-question.
19

c. #Subject-G Object verb
d. . . .
(27) What did Mary do afterward?
a. (Potom) Maruška || zavolala nějakého chlapce ze sousedství.
afterward Mary.Nom called some boy.Acc from neighborhood
b. #(Potom) Maruška nějakého chlapce ze sousedství zavolala
afterward Mary.Nom some boy.Acc from neighborhood called
‘Afterward Mary called some boy from her neighborhood.’
I have proposed in 1.1 that G-movement is a last resort operation. We predict that G-
movement takes place only if there is a non-G element asymmetrically c-commanding the
given subject. Since there is no such element, in this particular case there is no G-movement
taking place. A corresponding tree representation is given in (28).
20

(28) Derivation of [Subject]-G verb Object
vP
Subject G
vP
v
VP
t V
Object
Let’s now consider the same sentence but with both the subject and the verb given. An
example is given in (29) and (30).
(29) a. Subject-G verb-G Object ←−
b. #Object verb-G Subject-G
c. #Subject-G Object-G verb
d. . . .
(30) Whom did Mary call afterward?
a. Potom Maruška zavolala || nějakého chlapce ze sousedství.
then Mary.Nom called some boy.Acc from neighborhoud
b. #Potom Maruška nějakého chlapce ze sousedství zavolala
then Mary.Nom some boy.Acc from neighborhoud called
‘Then Mary called some boy from her neighborhoud.’
Since neither the subject nor the verb is asymmetrically c-commanded by anything new, it
follows that the derivation should be the same as in (28). The reason is that in neither (26)
nor (29) does G-movement take place. The representation I argue for is given in (31).
21

(31) Derivation of [Subject]-G verb-G Object
vP
Subject G
vP
v-G
VP
t V
Object
The logic that arises is the following: if a structure is monotonous in the sense that there is
no point where a new element would asymmetrically c-command a given element, syntax
does not have any tool to mark what part exactly is given and what part exactly is new. In
contrast, if there is any deviance from the basic word order, the partition between given
and new is syntactically realized. Thus, the interpretation of the utterance is restricted by
syntactic tools. I will argue in chapter 4 that if there is no G-movement, the partition is
established by the semantic component.
I argue that any basic word order sentence – if presented out of the blue – is ambiguous
with respect to its information structure. Since there is no syntactic marking of the partition
between given and new within basic word order determining the partition is left entirely to
the semantic interface. I argue that this follows from the fact that in a neutral word order
sentence there is no G-movement taking place. Thus the syntactic output of such a sentence
is identical no matter how many semantic interpretations of the sentence are available.
This conclusion is supported by the fact that there is no difference in prosody between
(27-a) and (30-a). If we assume that phonology reads prosody directly off the syntactic
structure (see, for example, Bresnan 1972; Truckenbrodt 1995; Wagner 2005; Büring 2006,
among many others) this is an unsurprising result. Therefore, it is only the semantic component
that may interpret such a clause in different ways depending on the actual context.
For syntax and the syntax-phonology interface there is only one structure to be considered.
Crucially, this behavior is a consequence of G-movement being a last resort operation. If G-
movement had to take place whenever there was something potentially given, the semantic
(but also word order and prosodic) ambiguity of the basic word order would be unexpected.
In this section, we have seen that there is a close connection between the last resort
character of G-movement and the multiple interpretations available for basic word orders.
In the next section we will look at ambiguities that arise within derived orders. I will argue
that this type of ambiguity follows from G-movement being dependent on head movement.
1.3 Verb partitions and their semantic ambiguity
In Czech the partition between given and new is often manifested by a finite verb. The
fact that Czech verbs usually appear between the given and the new part has already been
observed by Vilém Mathesius (Mathesius, [1929] 1983, 1939). In the following Czech
22

functionalist tradition there has been prevailing disagreement on whether the inflected verbal
form can be characterized as given or new (for example, Sgall (1967); Hajičová (1974);
Sgall et al. (1980)) or whether the verb forms a special category which is neither given nor
new (for example, Firbas 1964; Svoboda 1984). I will argue that the finite verb is indeed either
given or new. The reason why it is so difficult to characterize its information structure
status is that the verb is often ambiguous between being given and new. More precisely,
the partition between given and new often either immediately precedes the finite verb or
immediately follows it.
I will argue that this is a side-product of an independent property of G-movement. In
particular, I will argue that G-movement is parasitic on head movement. The idea that
a certain type of movement may be dependent on head movement has been already proposed
in other contexts (see, for example, Holmberg 1986 for Object shift and Heycock
and Kroch 1993; Johnson 2002 for coordination) but it is not well understood. I will provide
a possible motivation of this restriction on G-movement in section 2.3. In this section
I will show that we can find two possible motivations for head movement in connection
with G-movement. Either (i) head movement is an independent instance of G-movement,
and the verb is given, or (ii) the verb is new and head movement facilitates G-movement
of some other element. In both cases head movement is understood as part of the narrow
syntax (contrary to Chomsky (2000, 2001, 2004a)).
In the simple cases we have considered so far, the relative word order of the new part
was the same as the basic word order. Consider (32).
(32) a. And what about the lollipop?
b. #Lízátko malá holčička našla.←− # O || S V
lollipop.Acc little girl.Nom found
c. Lízátko || našla malá holčička. ←− O || V S
lollipop.Acc found little girl.Nom
‘A little girl found the lollipop.’
Our system as it is set up now predicts that the word order of the new elements should not
change, thus, the verb should linearly follow the subject, as in (32-b). This prediction is,
however, incorrect. As we see in (32-c), the verb must precede the subject. I argue that this
is a result of a more general restriction on G-movement that has not been discussed yet. A
first approximation of the restriction is given in (33). I will discuss the restriction in further
detail in section 2.3.
(33) Head movement restriction on G-movement:
α G can G-move out of XP only if X 0 moves out of XP as well.
The restriction is meant to capture the fact that if α G G-moves, the head in which projection
α G was base generated moves as well. We will see only in section 2.3 how exactly
the head movement restriction is motivated and when it applies. For now, let’s stay with
the following claim: there are two cases in which a head moves because of G-movement:
(i) the head itself is given and as such it needs to undergo G-movement; (ii) the head itself
23

is not given but its movement is necessary to facilitate G-movement of some other given
element.
As a consequence, we predict that the verb may appear on either side of the partition
between given and new. 17 If the head itself undergoes G-movement, the partition follows
the head. 18 If the head moves in order to facilitate G-movement, the partition precedes the
head. Thus we predict the pattern to be as in (34).
(34) a. O V || S ←− head undergoes G-movement
b. O || V S ←− head facilitates G-movement
The pattern is indeed correct. Example (32-c) is not only a felicitous answer to the question
And what about the lollipop? What happened to the lollipop? but it is also a felicitous
answer to the question Who found the lollipop?, as in (35).
(35) a. Who found the lollipop?
b. Lízátko našla || malá holčička.
lollipop.Acc found little girl.Nom
‘A little girl found the lollipop.’
In other words, the semantic ambiguity that we found with a basic word order repeats itself
on a smaller scale here as well. Even though we know that because of G-movement of
the object, the object is given (translated as the lollipop), we cannot conclude anything
about the status of the verb. The reason is that the head might have undergone either G-
movement, or it might have moved only in order to facilitate G-movement. The suggested
derivation is given in (36). For now I leave open the question of the exact landing site of
G-movement. In the following graphs, the landing site is marked as ?P.
(36) a. VP
V
Object
b. vP
Subject
vP
v-V
VP
V
Object
17 Notice that the verb can be either the rightmost element in the given part, or it can follow the given part
but it can never intervene between two given elements. Thus if, for example, both the subject and the object
are given the resulting order is either SOV or OSV, depending on other factors such as topicalization.
18 I put aside for now what determines the relative order of the given elements. I will address this question
in chapter 3.
24

c. ?P
v-V
vP
Subject
vP
v-V
VP
V
Object
d. ?P
Object ?P
v-V
vP
Subject
vP
v-V
VP
V
Object
I argue that the only difference between the interpretation in which the verb is given and the
interpretation in which the verb is new is in the motivation for head movement. Thus, the
syntactic output for PF and LF is the same. For native speakers, OVS structures presented
out of the blue are ambiguous and there is also no difference in the prosody they could use
to disambiguate.
We have now seen that there are two configurations under which information structure
ambiguity arises: either no G-movement takes place, thus, no partition is established, or the
partition could be placed in more than one position because there is no way to distinguish
between G-movement of the verb and independently motivated verb movement. It is left
up to the semantic interface to decide according to the relevant context. In chapter 4 I will
present a semantic system which is designed to make this type of decision.
In order to obtain the ambiguity of the new versus given partition with respect to the
verb and in order to derive the correct word order, I have assumed that G-movement is
dependent on head movement. In the next section I will look closely at predictions this assumption
makes. I will present evidence that the domain in which G-movement takes place
is indeed independently restricted by restrictions on head movement. Thus, head movement
and G-movement share the same locality restrictions. This fact thus provides further
25

support in favor of the assumption that G-movement is dependent on head movement.
1.4 The boundedness of G-movement
In this section, I will explore the further predictions and consequences of restricting G-
movement by head movement, namely, differences in locality restrictions on G-movement
that depend on the movement properties of the relevant facilitating head.
First of all, since head movement is clause bounded, we expect G-movement to be
clause bounded as well. This is indeed correct. As we can see in the following examples,
given elements can move only to the edge of their own embedded clause, even if the matrix
clause introduces only new information.
(37) For a long time I didn’t know what was going on with Mary.
a. Ale pak mi bývalá spolužačka řekla, že Marie si vzala Petra.
but then me former classmate told that Marie.Nom REFL took Petr.Acc
‘But then a former classmate of mine told me that Marie got married to Petr.’
b. Ale pak mi bývalá spolužačka řekla, že Marii potkalo velké
but then me former classmate told that Marie.Acc met big
štěstí.
happiness
‘But then a former classmate of mine told me that Marie got extremely lucky.
(She won a lottery.)’
c. #Marii mi bývalá spolužačka řekla, že potkalo velké štěstí.
Marie.Acc me former classmate told that met big happiness
In this respect G-movement differs from both wh-movement and contrastive focus movement,
which are not clause bound, as can be seen in (38) and (39).
(38) Koho, že jsi říkala, že si Marie
whom.Acc that Aux-you said that REFL Marie.Nom
‘Whom did you say Marie is going to Mary?’
bude
will
brát?
take
(39) Pavla, jsem ti říkala, si bude brát Marie, ne Lucie.
Pavel.Acc Aux-I you.Dat said REFL will take Marie.Nom not Lucie.Nom
‘I’ve (already) told you that it is Pavel who is going to marry Marie not Lucie.’
If G-movement were only restricted to clause boundaries, one might come up with another
explanation. For example, one could argue that this follows from the fact that G-movement
is A-movement (see appendix A for details)sec:AMovement. If this were the right explanation,
we would not expect further locality restrictions. In contrast, if G-movement is
restricted by head movement, we predict that G-movement is also not possible out of infinitival
domains because the infinitive head cannot move out of the domain either. The
latter prediction is borne out, as can be seen in (40). The relevant verbal head is marked by
a box.
26

(40) What happened to the antique chair you got many years ago from Mary?
a. Můj bývalý partner se pokusil tu židli
my former partner REFL tried that chair
‘My ex-partner tried to burn the chair.’
b. Můj bývalý partner chtěl tu židli spálit .
my former partner wanted that chair burn.Inf
‘My ex-partner wanted to burn the chair.’
c. Můj bývalý partner dokázal tu židli spálit .
my former partner managed that chair burn.Inf
‘My ex-partner managed to burn the chair.’
d. #Tu židli můj bývalý partner dokázal spálit .
that chair my former partner managed burn.Inf
spálit .
burn.Inf
‘My ex-partner managed to burn the chair. (OK: As to the chair, my expartner
managed to burn it.)’
If we do not bind G-movement to head movement, this behavior is unexpected also because
other elements, for example clitics, can climb up from the infinitival domain as in (41) with
a reflexive tantum posadit se ‘sit down’.
(41) a. Petr se chtěl posadit .
Petr.Nom REFL wanted sit-down
‘Petr wanted to sit down.’
b. Petr se dokázal posadit
Petr.Nom REFL managed sit-down
‘Petr managed to sit down.’
.
If we assume that infinitives that allow clitic climbing are restructuring verbs (cf. Dotlačil
2004; Rezac 2005), then the infinitival restriction on G-movement cannot follow from presence
of a phase boundary. Whatever allows the clitic to get to the main clause, should be
able to get the given element out as well. 19 As we can see, however, this is not so. On the
other hand, if we allow G-movement of α G to take place only if the relevant head moves as
well, the difference between clitics and given material is predicted.
Furthermore, we predict that if a verbal head is selected by another verbal head (an
auxiliary), G-movement should be blocked as well. We find a useful minimal pair if we
compare sentences in the Present or the Past tense with sentences in the Future tense. The
main verb in the future tense, in contrast to other tenses where the verb moves out of VP,
stays in VP and the future auxiliary is base generated in vP (Veselovská, 2004; Kučerová,
2005). As we can see in (42), (43), and (44), the difference in the different head movement
properties, especially lack of head movement in the Future tense, propagates to the domain
in which an element can undergo G-movement. While in (42) and (43) the verbal head is
free to move and the given object can therefore G-move, in (44) the given object can only
19 Clitic climbing in Czech is a syntactic, not a phonological, process. This can be shown, for example, by
the fact that different clitics in Czech climb differently. See Dotlačil 2004 for more details.
27

immediately precede the main verb. Any other instance of G-movement is not possible. 20
(42) a. What happened to the book?
b. Tu knihu || dala Marie Petrovi. ←− Past
the book.Acc gave Marie.Nom Petr.Dat
‘Marie gave the book to Petr.’
(43) a. What is happening to the book?
b. Tu knihu || dává Marie Petrovi. ←− Present
the book.Acc gives Marie.Nom Petr.Dat
‘Marie gives the book to Petr.’
(44) a. What will happen to the book?
b. Marie bude tu knihu dávat
Marie.Nom will the book.Acc give.Inf
‘Marie will give the book to Peter.’
c. #Tu knihu bude Marie dávat
the book.Acc will Marie.Nom give.Inf
Petrovi. ←− Future
Petr.Dat
Petrovi.
Petr.Dat
‘Marie will give the book to Peter.’
(OK as ‘As to the book (in contrast to the violin), Marie will give it to Peter.)
Interestingly, we find differing morphology even within one tense – the past tense. While
there is no overt auxiliary for 3rd person, 1st and 2nd person are formed by a finite auxiliary
and a past participle. If G-movement depends on head movement we predict non-local G-
movement of a given object to be possible with a 3rd person subject but not with a 1st or a
2nd person subject. This prediction is indeed correct, as can be seen in (66) and (67).
(45) 3sg.: Non-local G-movement possible:
a. What happened to the boat that got demeged in the last storm?
b. Lod’ opravil jeden technik.
boat.Acc repaired one technician.Nom
‘A technician repaired the boat.’
(46) 1pl.: Only local G-movement:
a. What happened to the boat that got demaged in the last storm?
b. Jeden technik a já jsme lod’ opravili .
one technician.Nom and I Aux.1pl boat.Acc repaired
‘A technician and I repaired the boat.’
Tying G-movement to head movement makes an additional prediction. If there is no head
that can move, only very local G-movement should be possible. A case to consider is a
small clause. As we can see in (47), this prediction is indeed correct. As with infinitives,
in a small clause, a given element can undergo only very local G-movement. No further
20 These examples are vastly simplified. To keep the pairs minimal I use an iterative form in the present
tense. The reason is that the corresponding word does not have a periphrastic future. Another important point
is that (44-c) is a plausible structure but the object would have to be interpreted as a topic (As to the book,
Marie will give the book to Peter.).
28

movement is possible.
(47) a. Why does Peter look so happy?
b. Marie je na Petra pyšná
Marie.Nom is of Petr proud
‘Marie is found of Peter.’
c. #Na Petra je Marie pyšná.
of Petr is Marie proud
.
Another interesting pattern arises from interactions of G-movement and the EPP requirement
of T. Simplifying considerably, in Czech T attracts whatever is the structurally closest
element (for more details see Kučerová (2005)). Thus, if there is a subject in Spec,vP, T
attracts the subject. On the other hand, if there is no subject (Czech is a pro-drop language),
T may attract the verbal head occupying v. See the examples in (48) and (49), illustrating
the point.
(48) a. Marie koupila husu.
Marie.Nom bought goose.Acc
‘Marie bought a goose.’
b. [ TP Marie T [ vP t Marie bought goose ] ]
(49) a. Koupila husu.
bought goose.Acc
‘She bought a goose.’
b. [ TP bought T [ vP t bought goose ] ]
We can now make an interesting prediction. If a given element moves to the left edge of
vP, then it can be attracted by T even if T is realized as an auxiliary. Thus, we expect to
find a word order that superficially looks like a violation of the head movement constraint
on G-movement. This is correct, as can be seen in (51). Notice that there is nothing wrong
with moving a given element over an auxiliary. The given element just cannot undergo
G-movement. Other instances of movement are not a problem.
(50) Scenario: And what about all the money you inherited?
(51) Všechny peníze jsem poslala osamělým dětem. ←− Past
all money Aux.1sg sent lonely children.Dat
‘I sent all (the) money to lonely children.’
(52) Past tense:
a. VP level: the given object moves over V:
29

VP
money
VP
gave
VP
children
t money
b. V moves to v:
vP
gave
VP
money
VP
t gave
VP
children
t money
c. the object undergoes G-movement over v:
vP
money
vP
gave
VP
t money
VP
t gave
VP
children
t money
d. T-auxiliary is merged and probes for the closest element:
30

TP
Aux
vP
money
vP
gave
VP
t money
VP
t gave
VP
children
t money
e. the given object moves to T:
TP
money
TP
Aux
vP
t money
vP
gave
VP
t money
VP
t gave
VP
children
t money
In contrast, in the future tense there is no V-to-v movement. Thus, a given object cannot
reach the vP edge, as shown by the object in (54). Since the future auxiliary is base generated
in v and there is no overt subject, T attracts the auxiliary. The derivation is schematized
in (55).
(53) Scenario: And what about all the money you will inherit?
(54) Budu všechny peníze posílat osamělým dětem.←− Future
will.1sg all money send.Inf lonely children.Dat
31

‘I will send all (the) money to lonely children.’
(55) Future tense:
a. VP level: the given object moves over V:
VP
money
VP
give
VP
children
t money
b. v-auxiliary is merged and the object cannot G-move further:
vP
Aux
VP
money
VP
give
VP
children
t money
c. T is merged and probes for the closest element:
TP
T
vP
Aux
VP
money
VP
give
VP
children
t money
d. the auxiliary moves to T:
32

TP
Aux
TP
T
vP
t Aux
VP
money
VP
give
VP
children
t money
Notice that the sentences we have considered in this section crucially differ from the sentences
that originally motivated introducing G-movement into the system. The difference
is that if head movement is blocked, α G may be asymmetrically c-commanded by new
material. Recall the definition of G-movement from (11), repeated below as (56). Even
though we can identify what element must be given, we cannot establish a perfect partition
between given and new as we were able to do in the previous cases.
(56) G-Movement [version 1]
G-movement must take place
a. iff α G is asymmetrically c-commanded by a non-G element,
b. unless the movement is independently blocked.
The definition of G-movement given in (11)/(56) takes care of these cases with the clause
unless the movement is independently blocked. Thus, the pattern that arises can be described
as do as much as you can. But does this mean that anything goes? I will address
this question in the next section. I will show that if α G is required to G-move, α G must
move at least once.
In chapter 4, I will address the question why it is sometimes okay to move a given element
only locally even if there is a higher new element, while in other cases, non-local
movement is required. The distinction will follow from the way we will interpret givenness.
For now, let’s stay with the soft constraint formulation which requires a given element
to move across as many new elements as syntactically possible.
In this section, we have seen that tying G-movement to head movement makes correct
predictions since G-movement is able to move only as far as the relevant head can move. If
movement of the head is independently blocked (for example, by the head being selected
by another head), G-movement can only be very local.
33

1.5 The distribution of pronouns: G-movement as a last
Resort
In this section (i) I will provide independent evidence for the assumption that G-movement
is a last resort operation. (ii) I will also address the question of what happens if α G is
required to move but the movement is independently blocked by syntax. I will also introduce
an observation that elements that are given by their lexical entry, such as pronouns,
do not undergo G-movement. I will use the difference between lexically given items and
non-lexically given items to investigate when G-movement must take place.
I will first look at cases in which G-movement is independently blocked. In contrast to
the previously discussed cases where α G was able to locally G-move, we will now consider
cases in which α G cannot move at all. We will see that, in such a configuration, if α G is
asymmetrically c-commanded by a new element, α G must be realized by a pronoun, i.e.,
an element that comes marked as given from the lexicon. I will argue that if G-movement
of α is required, α must G-move at least once. On the other hand, as we will see, if α G
is trapped in a position that is not asymmetrically c-commanded by any new element, α G
does not need to be lexically given. This fact will provide independent evidence for the
assumption that G-movement is a last resort operation.
I will refine this analysis in chapter 4 where I will introduce a global comparison system
which will enforce lexical givenness in cases that would not be otherwise interpreted
as given. In this chapter, however, we will stay with the surface oriented generalization.
A question we have not asked yet is whether G-movement applies to all given elements.
The answer is no. G-movement does not apply to pronouns. To see this, consider the
example in (57). As we can see, there is a difference in the position of the same given
element depending on whether the element is realized as a full DP (Pavel) or as a pronoun
(ho/jeho, ‘him’). 21
(57) What do you know about Pavel?
a. Marie ho viděla na nádraží. ̌new > pronoun
Marie.Nom him.Acc saw on railway-station
b. #Marie Pavla viděla na nádraží. # new > DP
Marie.Nom Pavla.Acc saw on railway-station
21 Personal pronouns in Czech come in two flavors: weak pronouns, in this case ho, and strong pronouns,
in this case jeho (Cardinaletti and Starke, 1999). They differ in their syntactic distribution and interpretation.
Crucially, weak pronouns are excluded from the left edge of a prosodic constituent, including the sentence
initial position. The reason is that Czech requires main word stress to be aligned to the left edge of a prosodic
constituent (see, for example, Petr et al. 1986; Palková 1994; van der Hulst 1999) and since weak pronouns
cannot be stressed they are excluded from the left edge. If a pronoun needs to be, for instance, in the sentence
initial position, it must be realized as a strong pronoun. For further details about the distinction between weak
and strong pronouns see Cardinaletti and Starke 1999. More details about distributional properties of Czech
weak pronouns can be found at Vos and Veselovská 1999.
34

c. #Jeho viděla Marie na nádraží. # pronoun new t
him.Acc saw Marie.Nom on railway-station
d. Pavla viděla Marie na nádraží. ̌DP new t
Pavel.Acc saw Marie.Nom on railway-station
‘Marie saw him/Pavel in the railway-station.’
One could argue that pronouns are excluded from the sentence initial position on independent
grounds. As we can see in (58), with a different information structure, i.e, if the
pronoun is contrastive, the pronoun initial clause is fully acceptable.
(58) JEHO Marie neviděla. Jenom Petra.
him.Acc Marie.Nom not-saw only Petr.Acc
‘Marie didn’t see HIM. She saw only Peter.’
It is well established that pronouns undergo movement from their base generated position.
22 I argue, however, that this is not G-movement. If pronouns underwent G-movement
we would expect them to appear in the same position as full DPs that are given. Most importantly,
they would not be able to be asymmetrically c-commanded by a new element. As
can be seen in (57-a), no such requirement holds for pronouns. In this particular example,
ho ‘him’ is asymmetrically c-commanded by Marie, a new element, but this is not a reason
for the pronoun to move. I argue that pronouns do not undergo G-movement because
G-movement is a last resort operation and as such it takes place if and only if the relevant
semantic interpretation would not be otherwise available. Since pronouns are already
marked from the lexicon as given, G-movement is not needed, hence not allowed. As such,
we can understand G-movement as a disambiguating method: G-movement applies only
to syntactic elements that are lexically ambiguous, i.e., G-movement applies only to lexical
items that can be interpreted either as given, or as new.
This brings up an important point: what is the purpose of G-movement? If G-movement
is a last resort operation, it cannot be the grammatical operation which marks or licenses
an element as given. And yet, it does not apply to an element which is already given. I will
argue in chapter 4 that the purpose of G-movement is to create a syntactic configuration
that can be interpreted by the semantic component as containing given elements. If an element
is already in the right configuration, G-movement does not apply to it. If an element
is already given, G-movement does not take place either.
At this point we do not have the right tools to explain the interactions between given
elements and G-movement. Before we can get there, we need to understand better the
properties of the syntactic configuration in which G-movement occurs.
So far our system does not contain anything that would enforce the difference between
elements marked as given from the lexicon and elements marked as given by the syntax. I
22 Czech weak pronouns are usually analyzed as second position clitics. As such they move to a position
following the first syntactic constituent. For details see, for example, Veselovská 1995; Franks and King
2000; Bošković 2001.
35

will thus modify the definition of G-movement from (11) in order to capture the distinction
between lexically given and lexically ambiguous items. The new definition of G-movement
is given in (59). A definition of what it means to be lexically given is in (60).
(59) G-Movement [version 2]
G-movement is required iff α G is asymmetrically c-commanded by a non-G element;
unless
a. the movement is independently blocked, or
b. α G is lexically given.
(60) Lexically given
α G is lexically given if its semantic lexical entry requires α G to have a salient
antecedent A such that α and A corefer.
Let’s first have a look at what happens if α G is required to move but it is in a position
from which it cannot move at all. 23 In Czech, G-movement is impossible out of a specifier
or out of a syntactic island. 24 Consider the context in (61). As we can see in (62)–(64), if
a given element is within a specifier, it cannot be realized as a bare NP/DP, as can be seen
in (a)–(b). 25 To make such a sentence felicitous, the given element must be realized either
as a possessive pronoun, as in the (c) examples, or the relevant DP must be modified by a
demonstrative pronoun, as in the (d) examples.
(61) Na programu byla diskuse o nové učitelce. ←− context I
on program was discussion about new teacher
‘The topic of the program was a discussion about a new teacher.’
(62) a. #Důvodem bylo podezření, že dcera (nové) učitelky bere drogy.
reason was suspicion that daughter of-new teacher takes drugs
‘The reason was a suspicion that a daughter of a new teacher is a drug-addict.’
b. #Důvodem bylo podezření, že učitelčina dcera bere drogy.
reason was suspicion that teacher’s daughter takes drugs
‘The reason was a suspicion that a daughter of a new teacher is a drug-addict.’
c. Důvodem bylo podezření, že její dcera bere drogy.
reason was suspicion that her daughter takes drugs
‘The reason was a suspicion that her daughter is a drug-addict.’
d. Důvodem bylo podezření, že dcera té (nové) učitelky bere drogy.
reason was suspicion that daughter of-that new teacher takes drugs
‘The reason was a suspicion that a daughter of that new teacher is a drugaddict.’
23 I thank to Danny Fox for suggesting this test to me and helping me to figure out its consequences.
24 It has been argued that Slavic languages, including Czech, allow Left-branch extraction, thus, movement
out of the specifier might be possible, see for example, Bošković 2005; Citko 2006. Putting aside whether or
not there is Left-branch extraction in Czech, the restriction on moving from specifiers does not need to follow
from this constraint. Since no head movement can take place out of a specifier, G-movement is expected to
be blocked independently of the Left-branch extraction properties of Czech.
25 In the (a) examples, the DP is realized post-nominally, while in the (b) examples it is realized prenominally.
Both options are possible in Czech. The difference is that a DP in a pre-nominal position cannot
be further modified.
36

(63) a. #Diskuse proběhla bez vědomí (nové) učitelky.
discussion ran without knowledge of-new teacher
‘The discussion went without knowledge of a new teacher.’
b. #Diskuse proběhla bez učitelčina vědomí .
discussion ran without teacher’s knowledge
‘The discussion went without knowledge of a teacher.’
c. Diskuse proběhla bez jejího vědomí.
discussion ran without her knowledge
‘The discussion went without her knowledge.’
d. Diskuse proběhla bez vědomí té (nové) učitelky.
discussion ran without knowledge of-the new teacher
‘The discussion went without knowledge of the new teacher.’
(64) a. #Nikdo nebyl na straně učitelky.
no-one not-was at side of-teacher
‘No one was on the side of a teacher.’
b. #Nikdo nebyl na učitelčině straně.
no-one not-was at teacher’s side
‘No one was on side of a teacher.’
c. Nikdo nebyl na její straně.
no-one not-was at her side
‘No one was on her side.’
d. Nikdo nebyl na straně té učitelky.
no-one not-was at side of-the teacher
‘No one was on the side of the teacher.’
As we can see in (65) and (66), other possible continuations of the context in (61), the
same observation holds also for islands. If an element that needs to undergo G-movement
is trapped in a syntactic island, then such an element must be realized either as a personal
pronoun, or it must be modified by a demonstrative pronoun.
(65) a. #Debata proběhla bez učitelky.
debate ran without teacher
‘The debate took place without a teacher.’
b. Debata proběhla bez ní.
debate ran without her
‘The debate took place without her.’
c. Debata proběhla bez této učitelky.
debate ran without this teacher
‘The debate took place without this teacher.’
(66) a. Bylo to myšleno jako útok proti ní
was it though as attack against her
‘It was meant as an attack against her.’
b. #Bylo to myšleno jako útok proti učitelce
was it though as attack against teacher
‘It was meant as an attack against a teacher.’
37

c. Bylo to myšleno jako útok proti této učitelce
was it though as attack against teacher
‘It was meant as an attack against this teacher.’
Consider now the example in (67). The example is offered here as a control showing that
in general there is no problem with having the unmodified form teacher as a continuation
of the discourse in (61). Crucially, in this example, G-movement is not required since there
is no new element asymmetrically c-commanding the DP učitelka ‘teacher’.
(67) Učitelka o diskusi (ale) nevěděla.
teacher about discussion but not-knew
‘However, the teacher did not know about the discussion.’
I argue that the reason why the non-modified DP in (62)–(66) is infelicitous as a continuation
of the context given in (61) is that the DP needs to be interpreted as given but it is in a
syntactic position from which it cannot undergo G-movement. Since there is no syntactic
operation that could achieve the desired semantic interpretation, inserting an element that
is lexically marked as given is the only option. 26 The generalization stating the relation
between pronouns and G-movement is given in (68).
(68) Generalization about syntactic distribution of pronouns
If α G is asymmetrically c-commanded by a new element and α G cannot move at
least once, then α G must be lexically given.
The above pattern provides an independent argument that G-movement is a last resort operation.
If there is no new element asymmetrically c-commanding α G , α G may be realized
as a full DP even if it is in a position from which it cannot move. It follows that an element
can be interpreted as given even if it does not undergo G-movement. Thus, we can
conclude that the purpose of G-movement cannot be to mark or license an element as given
(for example, by checking a G(iven) feature). There must be other tools to interpret an
element as given. In chapter 4 I will argue that the syntax-semantics interface does exactly
that.
If G-movement is a last resort operation, we predict that the restriction on specifiers
observed in (62)–(64) should disappear if the relevant specifier is the highest element in the
structure. As we can see in (69) this is indeed correct.
(69) Učitelovi žáci si totiž stěžovali na jeho učební metody.
teacher’s pupils REFL PART complained at his teaching methods
‘Students of the teacher complained about his teaching methods.’
Additional support for the argument that G-movement is a last resort operation comes
from coordinated DPs. If we assume the coordinate structure constraint (Ross, 1967), we
predict that a DP should not be able to move out from a coordination. Thus, a coordinated
given DP should be degraded, unless it is modified by a demonstrative pronoun. As we can
see in (70), this prediction is borne out.
26 Another option would be to choose a structure which allows G-movement to take place.
38

(70) a. #Žáky a učitelku to překvapilo.
students and teacher it surprised
‘Students and a teacher were surprised by that.’
b. Žáky a tu učitelku to překvapilo.
students and that teacher it surprised
‘Students and that teacher were surprised by that.’
c. Žáky i ji to překvapilo.
students and her it surprised
‘She (=the teacher) and students were surprised by that.’
At the same time we predict that, if the given element ‘teacher’ is the first conjunct, there
should be no need for G-movement to take place because according to the definition of G-
movement in (11), G-movement takes place only if there is a new element asymmetrically
c-commanding the given DP. Thus, if the given DP is independently high enough, it should
not matter that it is caught within a coordination. As we can see in (71), this prediction is
borne out as well.
(71) Na
on
programu
program
byla
was
debata o novém učiteli. . . .
debate about new teacher.Masc
(72) Učitele a (jeho) žáky to překvapilo.
teacher and his students it surprised
‘A teacher and (his) students were surprised by it.’
The examples in (73) are here as a control. If the same coordination is in an object position,
i.e., if it is further embedded in the structure, the fact that the given DP is the first conjunct
is of no help.
(73) a. #To se nelíbilo ani učiteli ani žákům.
it REFL not-liked nor teacher nor students
‘Neither a teacher nor students were happy about it.’
b. To se nelíbilo ani jemu ani žákům.
it REFL not-liked nor him nor students
‘Neither he nor students were happy about it.’
c. To se nelíbilo ani tomu učiteli ani žákům.
it REFL not-liked nor that teacher nor students
‘Neither the/that teacher nor students were happy about it.’
So far so good. We have seen an independent argument for the assumption that G-movement
is a last resort operation. At the same time we have seen that if α G is required to G-move,
α G must G-move at least once. Thus, we can formulate a minimal requirement on G-
movement, as in (74).
(74) Minimal requirement on G-movement:
If α G is asymmetrically c-commanded by a new element, α G must move at least
once. Otherwise, α G must be lexically given.
Immediate questions that arise are (i) what happens to an α G that does not have a lexically
39

given counterpart, and (ii) why is one instance of G-movement sometimes sufficient but
more than one instance of G-movement is required other times. I will leave the questions
open for now and I will come back to them in chapter 4.
In this section, I have provided two arguments that G-movement is a last resort syntactic
operation. The arguments were based on two facts about the distribution of pronouns in
Czech. First, I have shown that pronouns do not undergo G-movement. I have suggested
that this shows that G-movement applies only to elements that are to be interpreted as given
but that at the same time do not enter the derivation lexically marked as given. The second
part of this section addressed the question of what happens if an element cannot undergo G-
movement because of independent syntactic restrictions on movement, such as islands. We
have observed that in such configurations an element must already come from the lexicon
marked as given because there is no other way it could be marked as given in syntax. Last,
we have seen that if an element is trapped within an island but high enough in the structure,
the fact that it cannot undergo movement does not matter because the relevant interpretation
can be obtained without movement. Thus G-movement is not required.
1.6 Summary
In this chapter, I have offered several generalizations about the distribution of given and
new information in a Czech clause. I have shown that in an optimal configuration, the
given part linearly precedes the new part and the partition is marked by a verb which can
be on either side of the partition. I have proposed to account for the linear order facts in
terms of G-movement: an economy restricted movement that requires a given element to
move only if such an element is asymmetrically c-commanded by a new element and is not
lexically marked as given. I have presented several arguments in favor of understanding G-
movement as a last resort operation, including arguments from the distribution of pronouns
and the semantic ambiguity of certain strings. We have also seen that elements that are
lexically given do not undergo G-movement and that the linear partition between given and
new is not always perfect. In chapters 2 and 3 I will develop the current toy system in detail.
After that, in chapter 4, I will refine my answer to the question of how G-movement is
motivated and how it is treated by the interfaces. In particular, I will argue that G-movement
is a free syntactic movement, the purpose of which is to create a configuration which can be
semantically interpreted in a way that is required by the contextually established common
ground.
40

Chapter 2
G-movement
In chapter 1, I introduced several basic generalizations about Czech word order and its dependence
on the information structure of an utterance. I proposed to account for the Czech
word order facts by movement which takes place only if a given element is asymmetrically
c-commanded by a new element. I called this movement G-movement. I provided
evidence that G-movement is restricted by head movement. In particular, a given element
may G-move only if the head in the projection of which the given element was base generated
moves as well. Another important point I made was that G-movement cannot be
the grammatical tool that marks or licenses elements as given. The evidence comes from
the fact that only some given elements undergo G-movement. If a given element is in a
configuration where there is no asymmetrically c-commanding new element, the given element
not only does not move, it cannot move. A consequence of this observation is that
G-movement does not seem to be feature-driven.
The purpose of this chapter is to formalize the notion of G-movement and to fill in parts
of the syntactic system that are still missing. In particular, I will look closely at the following
three questions: (i) what is the syntactic target of G-movement? (ii) at what point of
the derivation may G-movement take place? and (iii) what kind of syntactic operation is
G-movement?
In section 2.1, I will provide a partial answer to the question of what a plausible landing
site of G-movement might be. I will argue that Czech has no unique syntactic position that
is always interpreted as given. Thus, we cannot associate G-movement with any particular
syntactic projection. I will refine the analysis of possible landing sites of G-movement in
section 2.2. I will concentrate on the question of when during the derivation G-movement
takes place. I will argue that G-movement may take place after any merge (both internal
and external), thus strengthening the point that G-movement may target various syntactic
positions. In section 2.3, I will explain the apparent lack of a unique landing site of G-
movement and the fact that G-movement may take place after any merge as a consequence
of G-movement being parasitic on head movement. I will show how the head movement
restriction relates to the timing of G-movement and to the apparent lack of a landing site
for G-movement. In particular, I will argue that the parasitic nature of this type of merge
is closely connected to the strict locality conditions and to the lack of a unique syntactic
41

landing site. In chapter 3, I will use the formalization offered in this chapter to show
how G-movement, a relatively simple syntactic tool, can account for complex word order
patterns.
2.1 The target of G-movement
In this section, I will address the question of the landing site of G-movement. I will show
that in Czech there is no unique syntactic position that is always interpreted as given. The
argument will be based on properties of movement to T. The logic behind this move is that
the element that moves to Spec,TP or to T is often interpreted as given. In the same time
such an element can be new as well. If, for example, Spec,TP were a unique syntactic
position for being interpreted as given, the possibility of new elements in Spec,TP would
be unexpected. 1 I will argue that movement to Spec,TP is a result of an Attract Closest
condition on T and as such it has nothing to do with the information structure nature of the
projection. I suggest that the ambiguous property of movement to Spec,TP can be extended
to other functional projections as well. 2
Notice that the argument goes only in one direction. It says that it cannot be true that
whatever moves to Spec,TP must be given. Thus, the argument does not exclude the possibility
that whatever is to be interpreted as given moves to Spec,TP. We have seen, however,
in section 1.4 that this is not true either. In section 2.2, I will provide examples that make
the point that G-movement can target other positions as well.
We now proceed to the first half of the argument, i.e., to the claim that Spec,TP does not
have a uniform interpretation. We have already seen in the discussion of basic word orders
(in section 1.2), that in the basic word order the leftmost element can be either given or new.
Consider the example in (1). Nějaká paní ‘some lady’ does not need to be presupposed and
it can still occupy the subject position (Spec,TP).
(1) a. What happened?
b. Nějaká paní poslala osamělým dětem dárky.
some lady.Nom sent lonely children.Dat presents.Acc
‘Some lady sent presents to lonely children.’
I argue that movement to T (Spec,TP or head-adjunction to T) in Czech is driven by an
EPP-like principle, not by, for example, some given feature (see also Saito (1989); Tateishi
(1994); Sauerland (1999) for scrambling in Japanese). If this is correct, we predict that
movement to Spec,TP is dependent on the internal structure of vP, in the sense that it is the
highest element within vP that moves to Spec,TP by virtue of being closest. Thus, even in
1 The argument is in line with Lavine and Freidin (2002); Bailyn (2003), among others, who argue that
Spec,TP is not a designated given position.
2 The argument crucially assumes that given DPs target an A-position. One might argue that the given
interpretation arises only if the relevant DP undergoes A-bar movement. I will argue in the appendix that
G-movement is indeed A-movement. Thus, there is no reason to assume that there is a positional difference
between a DP being interpreted as given and a DP being interpreted as new.
42

asic word order cases we expect word order variations depending on the internal structure
of vP.
We have already seen such a case in the discussion of unergative and unaccusative verbs
in section 1.1. Consider again the examples in (2) and (3). As these examples show, there
is a word order difference between unergative and unaccusative verbs. While an unergative
verb follows the subject, an unaccusative verb linearly precedes the subject.
(2) Basic word order of unergative verbs:
a. [Marie tancovala] New
Marie danced
b. #[Tancovala Marie] New
danced Marie
’Marie danced.’
(3) Basic word order of unaccusative verbs:
a. [Přijel vlak] New
arrived train
b. #[Vlak přijel] New
train arrived
‘A train arrived.’
If the subject of an unaccusative verb is structurally lower (VP) than the subject of an
unergative verb (vP), the word order difference follows under the following assumptions.
First, an inflected verb must undergo movement from V to v. 3 Second, in Czech a DP
can establish syntactic relations by Agree; it therefore does not need to raise from its base
generated position. Third, v does not need to have a specifier. The resulting structures are
given in (4) and (5). In the case of unergatives, T attracts the subject because it is the closest
element. On the other hand, in the case of unaccusatives, T attracts the v head (or the v-V
complex) because it is the structurally closest element. 4
3 In Czech, this condition holds both for finite and inflected infinitival forms such as participles. For
more details on syntax of Czech verbal morphology see, for example, Junghanns (1999); Veselovská (2004);
Veselovská and Karlík (2004); Kučerová (2005).
4 In functionalist Czech literature, verbs like ‘arrive’ are traditionally classified as ‘verbs of appearing on
the scene’ (Sgall et al., 1980, 1986, among others). Putting aside that it is unclear how to determine this
category, other unaccusative verbs behave in the same way even when they are not semantically verbs of
appearing on the scene (David Pesetsky, p.c.). See the examples in (i) and (ii).
(i) a. [Líbí se mi knihy] New
like REFL to-me books.Nom
b. #[Knihy se mi líbí] New
books.Nom REFL to-me like
‘Books appeal to me.’
(ii) a. [Existují různé názory na dějiny] New
exist various opinions.Nom at history
b. #[Různé názory na dějiny existují] New
various opinions.Nm at history exist
43

(4) Derivation of basic word order of unergatives:
TP
T
vP
subject
vP
v-V
VP
t V
(5) Derivation of basic word order of unaccusatives:
TP
T
vP
v-V
VP
t V
subject
I argue that there is nothing special about unaccusatives and unergatives with respect
to G-movement. The leftmost element can also be interpreted as given, without any word
order change, as seen in (6) and (7).
(6) a. What did Mary do?
b. Marie || tancovala
Marie danced
’Marie danced.’
(7) a. What arrived?
b. Přijel
arrived
|| vlak
train
‘A train arrived.’
Furthermore, as the examples in (8) and (9) show, if the word order is reversed, the all
new interpretation is not available anymore and the leftmost element must be interpreted as
given.
(8) a. What happened?
b. #[Tancovala Marie] New
danced Marie
’Marie danced.’
c. Who danced?
d. Tancovala || Marie
danced Marie
’Marie danced.’
‘There are various opinions about history.’
44

(9) a. What happened?
b. #[Vlak přijel] New
train arrived
‘A train arrived.’
c. What about the train?
d. Vlak || přijel
train arrived
‘The train arrived.’
To conclude, we have seen that even though an element in Spec,TP is often interpreted
as given, there is nothing about this particular syntactic position that enforces given interpretation.
I argue that this conclusion can be extended to other syntactic positions as
well, i.e., there is no syntactic projection that is canonically associated in Czech with given
interpretation. As we will see in the next section, G-movement may take place after any
merge. This means that it can target various syntactic positions. One might conclude that
basically any syntactic position is a good enough landing site for G-movement. In section
2.3, I will show, however, that the possibilities of various landing sites are restricted by G-
movement being parasitic on head movement. In chapter 4, I will refine the generalization
even further. I will show that the landing sites are further restricted by their semantic type.
2.2 When in the derivation does G-movement apply?
In this section I will address the question of timing of G-movement. In order to distinguish
among different theories, I will look more closely at more complex word order facts. On
the basis of this data, I will argue that G-movement may take place after any merge (both
internal and external).
I will consider in turn three hypotheses: (i) G-movement may take place only at a
phasal level, (ii) G-movement may take place at the end of any maximal projection, and
(iii) G-movement may take place after any merge, the hypothesis I will argue for.
(10) Timing of G-movement:
G-movement may take place after any merge.
Consider first the examples in (11) and (12). The example in (11) shows the basic word
order in a ditransitive construction. The example in (12) differs only in a local word order
change: while in the basic word order, the indirect object Pavel precedes the direct object
‘horse’, the order of the objects is reversed in (12). The rest of the order is the same.
(11) Neutral order: S V IO DO
a. What happened? / What did Mary give to Pavel? / . . .
b. (||) Marie (||) dala (||) Pavlovi (||) koně.
Marie.Nom gave to-Pavel.Dat horse.Acc
‘Marie gave Pavel a horse.’
(12) a. Whom did Marie give a horse?
45

. Marie dala koně || Pavlovi.
Marie.Nom gave horse.Acc Pavel.Dat
‘Marie gave a horse to Pavel.’
Crucially, I assume that the DO > IO order is not base generated. The argument is difficult
to make though. We will see, however, in chapter 3 that the phenomenon of local
G-movement reordering is widespread and rather complex; we can make better predictions
if we do not allow base generated word order permutations. I will also shortly comment on
base generation approaches in the appendix.
With this assumption in place, the question is how exactly we can derive the local word
order change in (12). Much depends on what we assume to be the position of the verb.
For example, if we assume that the verb may be higher than vP, for instance at TP, then all
three hypotheses would derive the same result. To see this point, consider the following
derivations.
(i) If G-movement takes place at the end of vP phase, we predict the following derivation.
5 After vP is built, the direct object (DO) undergoes G-movement to the edge of the
phase, resulting into (13-b). For convenience, I assume a Larsonian VP shell for ditransitive
verbs (Larson, 1988; Hale and Keyser, 2002). In the end of this section, I will address
the possibility that ditransitives are formed by an applicative head. In the next step of the
derivation, the verb and the subject move to T, as in (13-c). We have seen in 2.1 that in
Czech the subject does not need to move to Spec,TP but let’s assume for the sake of the
argument that it might move there. I put aside for now what exactly the landing site of
G-movement is. I mark the landing site as ?P.
(13) G-movement takes place at a phase level:
a. vP
subject
vP
v
VP
V
VP
IO
VP
V
DO
5 I assume that in Czech vP is a phase. It has been argued that VP can be phase in some languages as
well (cf. Ko (2007) for Korean and Fox and Pesetsky (2005) for Scandinavian languages). The difference is
immaterial for our discussion. The option that VP is a phase equals to the hypothesis that G-movement takes
place in the end of a maximal projection. As we will see shortly, this hypothesis make incorrect predictions
as well.
46

. ?P
DO
vP
subject
vP
v
VP
V
VP
IO
VP
V
DO
c. TP
subject
TP
T ?P
DO
vP
subject
vP
v
VP
V
VP
IO
VP
V
t DO
(ii) If G-movement takes place at the end of a maximal projection, the verb may stay in v or
it can move to T. The direct object moves to the edge of VP and stays there. The derivation
proceeds as in (14).
(14) G-movement takes place at a maximal projection level:
47

a. VP
DO
VP
V
VP
IO
VP
V
DO
b. vP
Subject
vP
v
VP
v
V
DO
VP
t V
VP
IO
VP
t V
t DO
(iii) If the movement takes place after any merge, the direct object can move immediately
after the indirect object is merged. 6 In the next step the verb moves higher, followed by
merge of the subject, as in (15).
(15) G-movement takes place after any merge:
a. ?P
DO
VP
IO
VP
V
DO
6 Notice that the direct object does not need to move over the verb because the verb is given as well.
48

. vP
subject
vP
v
VP
V ?P
DO
VP
IO
VP
Even though we cannot decide among the three hypotheses by looking at the example
in (12), the hypotheses make different predictions that can distinguish among them. The
first prediction concerns the position of the main verb. I have already mentioned in 1.4 that
in Czech the main verb obligatory moves out of VP in the present and the past tense but
not in the future tense (cf. Veselovská 2004; Kučerová 2005). Thus, if G-movement takes
place at the end of the phase or at the end of VP, the main verb in the future tense should
follow the direct object. In contrast, if G-movement takes place after any merge, the verb
is expected to be able to precede the direct object. The relevant structures are schematized
in (16).
(16) Different predictions for the position of the main verb in Future:
a. [ ?P DO [ vP subject v [ V P verb IO t DO ] ] ] ←− end of phase
b. [ vP subject v [ ?P DO [ V P verb IO t DO ] ] ] ←− maximal projection
c. [ vP subject v [ V P verb [ ?P DO [ V P IO t DO ] ] ] ] ←− any merge
V
DO
The examples in (17) show that only the hypothesis that G-movement takes place after
any merge makes the right prediction. 7 The example in (17-b), predicted by the two other
theories, is a plausible answer to the question What will Mary do with the horse?, i.e., it is
felicitous in a situation in which both the verb and the indirect object are new, but it cannot
be used when only the indirect object is new.
(17) Whom will Mary give a horse?
a. Marie bude [ V P dávat koně || Pavlovi].
Marie.Nom will give horse.Acc Pavel.Dat
b. #Marie bude koně dávat || Pavlovi.
Marie.Nom will horse.Acc give Pavel.Dat
7 The sentences in (17) may sound slightly awkward to a native speaker. The reason is that in order to keep
minimal pairs I use an iterative form of verb ‘give’ that is perfective. This is necessary because the basic root
does not have a periphrastic future. See also ftn. (41) in chapter 1, concerning the same issue.
49

‘Marie will give the/a horse to Pavel.’
Furthermore, the three hypotheses make different predictions with respect to the distribution
of adverbs. I assume that adverbs in Czech, with the exception of prosodically heavy
adverbs, are left branching and adjoin immediately to a functional head. Thus, for example
a vP adverb can linearly intervene between the subject and the finite verb in v. 8 Consider
the examples in (18). Time adverbials, such as yesterday, that are base generated at vP may
precede the finite verb and both the direct and indirect object, irrespective of the word order
of the objects. If G-movement of the direct object took place only at the phase level, this
order would be unexpected without additional assumptions about movement of adverbs.
Thus, we can conclude that the reordering of the direct and indirect object takes place
within vP.
(18) vP adverbs:
a. Marie [ vP včera dala Petrovi
Marie.Nom yesterday gave to-Peter.Dat
‘Marie gave yesterday Peter a horse.’
b. Marie [ vP včera dala koně
Marie.Nom yesterday gave horse.Acc
‘Marie gave yesterday a horse to Pavel.’
koně].
horse.Acc
Pavlovi].
Pavel.Dat
We can further restrict the domain of the reordering by looking at VP adverbs. Assuming
that VP adverbs adjoin to VP, we predict that if G-movement targets the end of the maximal
projection, the direct object should precede the adverb. If, on the other hand, G-movement
takes place within VP, there is no problem with the adverb preceding the direct object. As
we can see in (19), a manner adverb rychle ‘quickly’ may indeed precede the direct object.
(19) VP adverbs:
a. Marie dala [ V P rychle Petrovi
Marie.Nom gave quickly to-Peter.Dat
‘Marie gave quickly Peter a horse.’
b. Marie dala [ V P rychle koně
Marie.Nom gave quickly horse.Acc
‘Marie gave quickly a horse to Pavel.’
koně].
horse.Acc
|| Pavlovi].
Pavel.Dat
The examples in (20) are here as a control showing that the reordering with respect to the
VP adverb takes place within VP. Thus the only way we can derive the examples in (20)
with our current assumptions is to assume that G-movement takes place after any merge.
The derivation of (20-b) is schematized in (21). The only new element – the indirect object
– is marked by a box.
(20) vP and VP adverbs:
a. Marie [ vP včera dala [ V P rychle Petrovi koně]].
Marie.Nom yesterday gave quickly to-Peter.Dat horse.Acc
‘Marie gave yesterday quickly Peter a horse.’
8 For a detailed discussion of adverbial syntax in Czech, see, for example, Biskup (In preparation).
50

. Marie [ vP včera dala [ V P rychle koně
Marie.Nom yesterday gave quickly horse.Acc
‘Marie gave yesterday quickly a horse to Pavel.’
|| Pavlovi]].
Pavel.Dat
(21) a. VP is built and the DO undergoes G-movement above the IO:
?P
horse
VP
Pavel
VP
gave
t horse
b. V moves within the VP shell:
VP
gave ?P
horse
VP
Pavel
VP
t gave
horse
c. the VP adverb is merged:
VP
quickly
VP
gave ?P
horse
VP
Pavel
VP
d. the verb moves to v:
t gave
horse
51

vP
gave
VP
quickly
VP
t gave ?P
horse
VP
Pavel
VP
t gave
horse
e. the vP adverb is merged:
vP
yesterday
vP
gave
VP
quickly
VP
t gave ?P
horse
VP
Pavel
VP
f. the subject is merged:
t gave
horse
52

vP
Marie
vP
yesterday
vP
gave
VP
quickly
VP
t gave ?P
horse
VP
Pavel
VP
The argument as it stands now is not conclusive. First of all, if we assumed that a
ditransitive verb does not have a VP shell but there are instead two independent heads
involved (cf. for example, Pylkkänen (2002) and the literature cited there), the hypothesis
that G-movement takes place at the end of a maximal projection would still be feasible.
Similarly, if we assumed that adverbs form their own functional projections, we could not
distinguish between the two hypotheses. A crucial piece of evidence that G-movement may
take place after any merge comes from SOV orders. Consider the example in (22).
(22) a. How did the boy get the lollipop?
b. Chlapec lízátko
boy.Nom lollipop.Acc
|| našel.
found
‘The boy found the lollipop.’
In this case, the subject and the object are given and only the verb is new. Since the verb
needs to move to v, there must be an option for the object to move over the verb before the
subject is merged. If the object moved at the end of the vP projection, it should linearly
precede the subject, resulting in OSV order. As the example in (22) shows, this prediction
is incorrect. The corresponding derivation is given in (23). The given elements are marked
by boxes.
t gave
horse
(23) a. v is merged:
53

vP
found
VP
lollipop
VP
t found
t lollipop
b. the object moves:
vP
lollipop
vP
found
VP
t lollipop
VP
t found
t lollipop
c. the subject is merged:
vP
boy
vP
lollipop
vP
found
VP
t lollipop
VP
t found
t lollipop
The conclusion that G-movement takes place after any merge, of course, follows only
under the assumption that the object cannot tuck in under the subject (Richards, 1997). I do
not adopt this theoretical option here for two reasons. First, as we will see in chapter 3, it is
not always the case that if more than one given element moves, that the elements end up in
the same order. If we allowed tucking in, we would have to further restrict its application
only to a limited number of cases. Furthermore, tucking in in Czech always comes with a
special semantic interpretation. For example, in Czech (as well as in Russian and Polish), if
multiple wh-movement obeys superiority, the only possible answer is an answer with a pair
54

list reading. No such requirement exists for cases in which wh-elements move on nested
paths (see Meyer (2003, 2004) for more details). Consider the following examples.
(24) (Meyer, 2003, ex. (5))
a. After the film ‘Four weddings and a funeral’, Petr once more enumerated. . .
b. kdo si koho vzal na těch čtyřech svatbách.
who.Nom REFL whom.Acc took on these four weddings
‘married whom on the four weddings.’
(25) (Meyer, 2003, ex. (6))
a. There is going to be a wedding at our church tomorrow.
b. #Prosím tě, a kdo si koho bere?
beg you and who.Nom REFL who.Acc takes
‘Tell me who is going to marry whom.’
While the subject-wh ≻ object-wh order is acceptable in the situation in (24), in which all
the relevant brides and grooms have been introduced in the previous context (by watching
the movie) and the speaker asks for pairs formed by these brides and grooms, the same
order is not felicitous in the dialog in (25), in which the participants of the wedding are
unknown to the speaker. On the other hand, the reversed order is felicitous, as in (26).
(26) a. There is going to be a wedding at our church tomorrow.
b. Prosím tě, a koho si kdo bere v tuhle
beg you and who.Acc REFL who.Nom takes in this
‘Who is going to get married in this time of the year?’
roční
year
dobu?
time
I do not have any explanation for the relation between syntax and the semantic interpretation.
I state the correlation here only as an empirical observation. But we can still use it
as an argument against tucking in in other environments. In particular, I argue that tucking
in is not a strategy available for G-movement. 9 My main argument is that the SOV order
in (22) is not associated with any pair-list reading interpretation that we find elsewhere,
for example with contrastive focus. The SOV order simply asserts that both the boy and
the lollipop are given. The example could be used not only as an answer to the question
in (22) but would be felicitous in the following context as well: ‘A boy was sad because
he lost a lollipop that he got from his grandma. But then [the boy found the lollipop] and
he was happy again.’ Thus, since the object cannot tuck in under the subject, I argue that
the SOV order provides additional evidence for the hypothesis that G-movement may take
place after any merge.
I argue that we can generalize the lack of tucking in for G-movement. In particular, I
argue for the following condition on G-movement.
(27) Extension condition on G-movement:
G-movement must extend the tree.
9 I do not argue that there is no tucking in in syntax. My point is that in Czech multiple movement that
obeys tucking in obligatorily obtains a pair-list reading interpretation.
55

We can conclude that G-movement may take place after any merge. The reader might
object though that so far we have seen only cases involving verbal projections. As we will
see in the next section, this follows from the proposal that G-movement is parasitic on overt
head movement. Since there is no overt head movement in other environments, the properties
of G-movement in non-verbal environments cannot be directly tested. 10
To conclude, I have argued in section 2.1 that in Czech there is no unique syntactic position
that is interpreted as given. In this section I have made the argument even stronger,
by showing that G-movement may take place after any merge, and can thus target various
syntactic positions. The argument I presented here was based on the word order in ditransitive
sentences – in particular, on differences in the position of the main verb with respect to
the direct and the indirect object and then on the position of vP and VP adverbs with respect
to the direct and the indirect object. The conclusion that G-movement may take place after
any merge raises an immediate question: what exactly is the landing site of G-movement?
I will address this question in the next section.
2.3 The head movement restriction on G-movement
In this section, I will offer an explanation for the apparent lack of a unique landing site of
G-movement and for the fact that G-movement may take place after any merge. I will argue
that these two properties are a result of G-movement being parasitic on head movement.
I will motivate the head movement restriction on G-movement by a general condition on
merge, namely, merge dependence on agree. 11
In the previous section, we have seen that G-movement may take place after any merge.
We have also seen, in section 1.4, that G-movement of a G-element α forces head movement
of the relevant head. In the following discussion I will refer to the relevant head as
the head of α.
The picture that arises is the following: (i) If the G-element does not have a head
that could move, then G-movement is extremely local, as in (28). (ii) If the G-element
has a head that can move, G-movement can be non-local, as in (29). Notice that in both
cases, after G-movement takes place, the G-element is adjacent to the verb it was originally
merged with. The examples are schematized in (30).
(28) a. What will happen with the refugees?
b. Musíme doufat, že nějaká
must.1pl to-hope that some
nezisková
non-profit
organizace bude běžencům
organization will refugees.Dat
10 Veselovská (1998) argued that there is N-to-D movement in Czech. Her argument is based on fairly
complex possessive constructions including numerals. In these constructions we can indeed find local reorderings.
Unfortunately, I have not figured yet how to control for the involved information structure. I leave
the nominal cases for future research.
11 For concreteness I am assuming the general framework of Chomsky (2000, 2001, 2005). As far as I can
tell nothing much in the current proposal hinges on this choice, but I will not attempt here to explore the
compatibility of the proposal with alternative frameworks.
56

poskytovat jídlo.
provide food.Acc
‘We must hope that some non-profit organization will provide the refugees
with food.’
(29) a. What happened with the refugees?
b. Slyšel jsem, že běžencům
heard Aux.1sg that refugees
organizace jídlo .
food.Acc
poskytla
provided
(nakonec) nějaká nezisková
in-the-end non-profit organization
‘I heard that the refugees were provided with food by some non-benefit organization.’
(30) Dependence of G-movement on head movement:
a. organization will refugees provide with food
b. refugees provided organization with food
In order to account for the adjacency restriction observed in (28) and (29) (and in previous
examples in section 1.4), I will tie the adjacency requirement to independent properties
of agree and merge. In particular, I will follow Sigurðsson (2004) and Pesetsky and Torrego
(2006) in the proposal that any merge is preconditioned by agree, i.e., for merge to
take place, there must be feature matching involved. 12,13
(31) Agree Condition On Merge (Sigurðsson, 2004, ex. (11))
Two objects or elements, X and Y, may be merged only if the relation of Agree
holds between them.
I propose that the Agree Condition on Merge corresponds to two different feature configurations.
For an element α to move, there must be either (i) an externally merged element
β that brings new features into the derivation to be matched and/or valued; or (ii) α must
be re-merged within a projection that contain the head in which projection α was base generated.
It is the second case that interests us here. I argue that G-movement is this kind
of merge and that the head movement restriction on G-movement reflects such a feature
matching requirement. The head movement condition on G-movement is repeated below
as (32).
(32) Head movement restriction on G-movement:
12 The idea that external merge is feature-driven is not new. For example, Svenonius (1994); Collins (2002);
Adger (2003); Heck and Müller (2006) formulated such an idea in connection with subcategorization features.
13 The formulation of the conditon in Pesetsky and Torrego (2006) is the following:
(i) Vehicle requirement on Merge (VRM) (Pesetsky and Torrego, 2006, ex, (1))
If α and β merge, some feature F of α must probe F on β.
57

α G can G-move out of XP only if X 0 moves out of XP as well.
Thus, I argue that the head movement condition understood as a form of parasitic relation
(instead of new feature evaluation, use an already existing feature set) is a way to
create a landing site in case there is no externally merged probe available. In other words,
while in the case of feature-driven movement, there is probe P that creates a new sisterhood
relation (P lands itself as the landing site), in case of G-movement, there is no Probe and
re-merge is possible only if an existing feature-matching set is reused.
A consequence of this assumption is that a given element α must be remerged within
the same projection to which α’s head moves. We can thus strengthen the condition on the
head movement, as in (33).
(33) Head movement restriction on G-movement [final]:
a. α G can G-move out of XP only if X 0 moves out of XP as well.
b. If α G G-moved out of XP, α G may G-move to YP only if X 0 moves to YP as
well.
Another consequence of the feature matching requirement is that G-movement can take
place after any merge because it is not dependent on introducing new features into the
derivation. There is no need for G-movement to be feature driven. 14 Furthermore, since
head movement is very local, G-movement must be local as well.
Notice that there are three distinct configurations when G-movement of α can arise: (i)
the head of α is in its base generated position, as in (34); (ii) the head of α has undergone G-
movement, as in (35); (iii) the head of α underwent an independently motivated movement,
as in (36).
(34) Case (i): the sister of α is in its base generated position:
X
α
X
X
α
(35) Case (ii): the sister of α undergoes G-movement:
14 I will spell-out how exactly G-movement is driven in chapter 4. For the current discussion it is immaterial
whether G-movement is feature driven or is not. Notice though that if G-movement is not feature driven,
the proposal is still compatible with the phasal theory of Chomsky (2004b, 2005). Chomsky proposes that
a feature-driven operation may take place only at the phasal level because it is only the phasal head that
introduces the relevant feature(s) into the derivation. If G-movement is not feature driven, there is no reason
to wait for the phase to be completed. G-movement may take place at any point.
58

a. X
X
X
Z
X
X
α
b. X
α
X
X
X
Z
X
X
α
(36) Case (iii): the sister of α undergoes feature-driven movement:
a. X
X
Y
Y
Z
Z
X
X
α
b. X
α
X
X
Y
Y
Z
Z
X
X
α
We have already seen examples of all three configurations. Case (i) is a case of an object
moving locally around an infinitive, as in (37). Case (ii) corresponds to an OVS order if
59

oth the verb and the object are given, as in (38). Case (iii) demands, however, a more
careful discussion.
(37) Example of Case (i):
a. Pavel hasn’t come back home yet, right? Do you know what they plan to do?
b. Někdo snad bude Pavla hledat t Pavel .
someone.Nom hopefully will Pavel.Acc look-for
c. [ V P Pavel look-for Pavel ]
(38) Example of Case (ii):
a. Who ate the lollipop?
b. Lízátko snědla || Maruška t snědla t lízatko .
lollipop.Acc ate Maruška.Nom
‘Maruška ate the lollipop’
c. [ vP lollipop ate Maruška t v−V [ V P t V t lollipop ] ]
Notice that already the example in (38) is a case of a combination of independent movements
(V-to-v movement), i.e., an instance of Case (iii). Since both the verb and the object
are given, there is no need for G-movement within VP. It is only after V moves to v and the
subject is merged, when the need for G-movement arises. At this point, both the verb and
the object undergo G-movement but only the object moves from its base generated position.
There is one more instance of movement of the head of α that is relevant for our current
discussion. This is a case of head movement that happens in order to facilitate G-movement
(compare the discussion of (32) in section 1.3). A typical example is verb head movement
in an OVS where only the object is given and the verb and the subject are new. See example
(32) from section 1.3, repeated below as (39).
(39) a. And what about the lollipop?
b. Lízátko || našla malá holčička.
lollipop.Acc found little girl.Nom
‘A little girl found the lollipop.’
c. [ vP lollipop found little girl t v−V [ ?P t V t lollipop ] ]
Even though the derivation in (39-c) is reminiscent of the derivation in (38-c), the motivation
of the final verb movement seems to be rather different. I want to suggest that the
difference between head movement as an instance of G-movement and head movement as
a way to facilitate G-movement does not have any reflex in the syntax.
I propose that in both cases, i.e., both in OV||S and O||VS, the observed verb movement
is an instance of v-to-T movement. We have already seen that a finite verb in Czech does
not need to move overtly higher than v (in section 2.2). On the other hand, it is a plausible
assumption that there is Agree between T and v. I suggest that a verbal head may use the
60

existing Agree relation in case it allows an instance of G-movement. Furthermore, such
a head must use an existing Agree relation instead of extending or creating a projection.
Notice that this is a reversed side of the head movement restriction on G-movement. A
given XP can take a free ride on its head in the sense that it can use a set of existing
matching features with the head. But a head itself does not have such an option. For a head
to G-move, the head must enter into a new feature checking relation. v-to-T movement is
exactly such a case.
(40) v-to-T Movement in Czech
v-to-T movement in Czech takes place iff it changes the semantic interpretation.
(41) Restriction on overt head movement
If a head moves as an instance of G-movement or in order to facilitate G-movement,
the head must use an independently existing Agree relation.
Since there is no obligatory v-to-T movement in Czech, I argue that such movement is
possible only if it gives rise to a new semantic interpretation. We will see in chapter 4 that
we can unify the two types of head movement – head movement which is G-movement
and head movement that facilitates G-movement – even further. The point will be that the
purpose of any G-movement is to create a syntactic configuration that allows an element α
to be interpreted as given in case α would not be otherwise interpreted as given. Whether
the element that G-moves is given or new will turn out to be immaterial.
With the assumption that v moves to T both in case the verb undergoes G-movement
and in case head movement of the verb facilitates G-movement, the derivation of O||VS and
OV||S becomes identical. 15 It is only up to the semantic interface to decide on which side
of the partition the verb is. The corresponding derivation is given in (42).
(42) Derivation of OVS order:
a. vP
subject
vP
v-V
VP
t V
object
15 This is not entirely correct. If the object is presupposed and the verb is new, the object would move over
the verb already in VP. I put this detail aside.
61

. TP
object
TP
T-v-V
vP
subject
vP
t v−V
VP
t V
t object
The alert reader has already noticed that the derivation given above follows only if we assume
that the object cannot G-move above the subject without the verb moving as well.
There is nothing in the head movement restriction on G-movement that would block a
derivation in which the object would move above the subject within vP. As we will see in
chapter 3, there is also no problem with having more than one element preceding the verb
(if all the elements are given). Why then must the verb move?
I want to suggest that it is an independent property of specifiers (subjects) that nothing
can be merged within XP once the subject is merged in Spec,XP. At this point I do not have
any explanation of this fact but it seems to be true for Czech. (Recall that, for example,
adverbs are merged between the subject and the finite verb, unlike in English.) Thus, once
a subject (Spec,XP) is merged, no further element can be (re-)merged within XP. For an
element to be re-merged, a new head must be merged. It follows that if α needs to be G-
moved within XP, it must be done before Spec,XP is merged. Once Spec,XP is merged, α
can G-move only if X moves to a higher head. 16
In this section, I have offered an explanation for the head movement restriction on G-
movement. The basic idea of this condition was that two elements can be re-merged if the
new merge relation preserves the feature matching set of the original merge relation. This
condition has been introduced in order to account for the fact that G-movement is restricted
by overt movement of its sister and that G-movement may take place after any merge. It
follows that G-movement is not tied to a particular syntactic position. The only restriction
is that a given element may be re-merged only in the same projection as its head. I have also
addressed the question of when and where a head moves in order to facilitate G-movement.
16 I crucially assume that head movement cannot extend a phrase. For a proposal suggesting that head
movement leads to phase extension see den Dikken (To Appear).
62

2.4 Summary
In this chapter, I have addressed the question of what kind of movement G-movement is. In
particular, I have looked closely at the following three questions: (i) what is the syntactic
target of G-movement? (ii) at what point of the derivation does G-movement take place?
and (iii) what kind of syntactic operation is G-movement?
In section 2.1, I showed that Czech has no unique syntactic position that is always interpreted
as given. In the next section, section 2.2, I provided further evidence in favor
of this claim by showing that G-movement may take place after any merge. I explained
these properties of G-movement by tying them to the observation that G-movement of α is
restricted by the head movement possibilities of the head of α. In section 2.3, I suggested
that an element α can move without a feature trigger if the new merge relation contains
the feature matching set of the original merge relation and if the new structure affects the
semantic interpretation.
In the next chapter, I will refine the system in place by looking at more complex derivations
that contain several instances of G-movement. In particular, I will look at utterances
where more than one element can precede the verbal partition. I will show that these strings
can be derived by several instances of movement or they can be derived by phrasal movement
of a whole subtree. The combination of these two strategies will allow us to capture
various word order combinations found in Czech.
63

Chapter 3
Complex Word Order Patterns
So far we have considered only simple cases of G-movement, i.e., cases where only one
given element undergoes G-movement. In this chapter I will closely examine more complex
cases, i.e., cases where there is more than one given element undergoing G-movement.
The question is whether the relatively simple tools we have developed so far can account
for the complex word order data in Czech. I will argue that G-movement and independent
restrictions in the Czech syntax are indeed all we need.
First, I will investigate the relative position of the relevant verbal head with respect to
given elements. We will see that the relevant verbal head always follows all the given elements.
Thus, the verbal head marks the partition between given and new. 1 In section 3.1, I
will address the verb partition generalization in the context of the head movement restriction
on G-movement. We will see that the verbal partition follows from this restriction.
In section 3.2, I will look closely at cases where there is more than one given element
that needs to undergo G-movement. I will argue that multiple given elements either (i)
move in separate instances of G-movement, or (ii) there is a constituent dominating several
given elements and the costituent moves. Then I will address the question of when
several given elements may move as one constituent dominating several given elements. I
will argue that several given elements can move as one constituent only if (i) the smallest
constituent dominating them can independently move, and (ii) the constituent contains only
given elements. If these two conditions are not satisfied, given elements must undergo individual
G-movement. As we will see, these two types of movement – one constituent versus
several separate movements – result in different word orders. If given elements move as
one constituent, their relative word order is preserved, if they move separately, their relative
word order is reversed. I will argue that it is a combination of these two moving strategies
(plus base generation) which derives the variety of word order patterns found in Czech.
Thus, G-movement will turn out to be a sufficient tool for the Czech data.
1 According to Zikánová (2006), in the Prague Dependency Corpus, version 2.0, 92,35 percent of finite
verbs creates the partition between ‘theme’ and ‘rheme’. The distinction between theme and rheme roughly
corresponds to our distinction between given and new. Zikánová does not provide further characteristics of
the Modern Czech data (her work concentrates on older stages of Czech).
64

3.1 Deriving the verb partition
In this section I will investigate the relation between the position of the finite verb and the
partition between given and new. We will see that if more than one element precedes the
verb, the pre-verbal elements must be given. 2 I will argue that this is a result of the head
movement condition on G-movement, repeated in (1).
(1) Head movement restriction on G-movement [final]:
a. α G can G-move out of XP only if X 0 moves out of XP as well.
b. If α G G-moved out of XP, α G may G-move to YP only if X 0 moves to YP as
well.
In other words, a given element may move only if it is remerged within a projection in
which its head is remerged. This follows from our assumption about conditions on merge
from section 2.3. Furthermore, if more than one given element moves and the elements
belong to the same head, they must all be remerged within the same projection. The next
section will investigate consequences of this assumption and will look closely at derivations
containing more than one G-movement per time.
Consider first ditransitive constructions where only one of the objects is given and the
rest of the clause is new. So far we have only seen that such an object can undergo short
movement (in section 1.4) but we have not seen how far the object can move and where
the verb is. As we can see in (2)–(3), the object must be the leftmost element and the finite
verb must immediately follow the fronted object. The examples in (4)–(5) exemplify the
pattern.
(2) a. DO-G V S IO
b. #DO-G S V IO
c. #DO-G S IO DO-G S V IO
(3) a. IO-G V S DO
b. #IO-G S V DO
c. #IO-G S DO V
(4) What happened to the new chair you got from Mary? I haven’t seen it around.
a. Tu novou židli || dala naše uklízečka
své známé.
that new chair.Acc gave our cleaning-woman.Nom REFL friend.Dat
b. #Tu novou židli || naše uklízečka
dala své známé.
that new chair.Acc our cleaning-woman.Nom gave REFL friend.Dat
c. #Tu novou židli || naše uklízečka
své známé dala.
that new chair.Acc our cleaning-woman.Nom REFL friend.Dat gave
‘Our cleaning woman gave the chair to her friend.’
(5) Do you know what happened to Pavel?
a. Pavlovi || dala jeho maminka kytaru.
Pavel.Dat gave his mother.Nom guitar.Acc
2 This generalization in fact covers also contrastive elements because they are given as well.
65

. #Pavlovi || jeho maminka dala kytaru.
Pavel.Dat his mother.Nom gave guitar.Acc
c. #Pavlovi || jeho maminka kytaru dala.
Pavel.Dat his mother.Nom guitar.Acc gave
‘Pavel’s mother gave him a guitar.’
The pattern observed in (2)–(3) is not a result of a V2 requirement. As we can see in (6)
and (7), 3 for the direct object case, and in (8) and (9), for the indirect object case, there is
nothing inherently wrong with the finite verb being in a different position. But the position
is crucially dependent on the relevant context. In all the examples, the finite verb must
immediately follow all of the given elements.
(6) Do you know what our cleaning lady did with the new chair we got from Mary?
a. Tu novou židli naše uklízečka
|| dala své známé.
that new chair.Acc our cleaning-woman.Nom gave REFL friend.Dat
‘Our cleaning woman gave the chair to her friend.’
(7) I have heard that a friend of your cleaning lady stole from her the new chair? Is it
true?
a. Naše uklízečka
své známé tu novou židli || dala.
our cleaning-woman.Nom REFL friend.Dat that new chair.Acc gave
‘(No, that’s not true.) Our cleaning woman GAVE the chair to her friend.’
(8) Do you know what Pavel’s mam did to him that he is now so happy?
a. Maminka Pavlovi || dala kytaru.
mother.Nom Pavel.Dat gave guitar.Acc
‘Pavel’s mother gave him a guitar.’
(9) Is it true that Pavel was forced by his mam to buy a quitar from his pocket money?
a. Maminka Pavlovi kytaru || dala.
Pavel.Dat mother.Nom guitar.Acc gave
‘(No.) Pavel got a guitar from his mum. (He did not need to buy it from his
pocket money.)’
The above examples exemplify the fact that the finite verb behaves as a partition between
given and new. The question is how exactly we can explain the relation between the position
of the finite verb and the partition between given and new. There are two logical
options: either (i) the given elements move to separate projections above the projection
of the finite verb, or (ii) the given elements are adjoined to the same head. It follows
from our new definition of the head movement constraint and the extension condition on
G-movement that all given elements must be adjoined to the projection hosting the finite
verb.
3 The examples in (6) and (7) have a different word order of the arguments than the examples in (4). The
word order difference is irrelevant for the argument: the same orders as in (4) are grammatical if contrastive
focus was involved. I simplify the case here to stay only in the domain of G-movement.
66

We will see in the next section how the system works. In the rest of this section, I will
continue looking at simple examples that we have not considered so far. The examples will
serve as background for the coming discussion of more complex cases. Consider again
the example in (4), i.e., a structure in which only the direct object is given. The predicted
derivation proceeds as follows. First, the direct object needs to move at the VP level because
it is asymmetrically commanded by the new verb, (10-a). Then the new indirect
object is merged and the direct object moves above it, (10-b). Now the verb moves and the
direct object moves again, (10-c). In the next step, the verb undergoes V-to-v movement
and the direct object is free to move before the subject is merged, (10-d).
In the next step, the subject is merged as the specifier, (10-e). At this point the direct
object is again asymmetrically c-commanded by a new element (the subject). But the direct
object cannot move at this point. This follows from our assumption discussed in section
2.3 that nothing can be merged within XP after a subject was merged as a specifier of XP.
Thus, the only way the derivation can proceed is if the verb moves first to T. Only then is
the direct object free to G-move again, as in (10-f).
(10) Derivation of DO-G V S IO
a. VP
DO
VP
V
t DO
b. VP
DO
VP
IO
VP
t DO
VP
V t DO
67

c. VP
DO
VP
V
VP
t DO
VP
IO
VP
t DO
VP
t V
t DO
d. vP
DO
vP
v
VP
v
V
DO
VP
V
VP
IO
VP
t DO
VP
t V
t DO
68

e. vP
subject
vP
DO
vP
v
VP
v
V
DO
VP
V
VP
IO
VP
t DO
VP
t V
t DO
f. TP
DO
TP
T-v-V
vP
Subject
vP
v
VP
v
V
DO
VP
t V
VP
IO
VP
t DO
VP
t V
t DO
So far so good. Let’s now look at a derivation of a clause where both the subject and
the direct object are given, as in (6). The final word order is S-DO-V-IO. The first steps of
the derivation are identical with (10-a)–(10-d). What happens after the subject is merged,
69

(10-e)? At this point all needs of the given elements are satisfied, i.e., there is no given element
that would be asymmetrically c-commanded by a non-presupposed element. Thus,
from the point of view of G-movement, the derivation is complete. 4
Even though the example in (4) contains two given elements, the derivation proceeds
without any interesting interaction between them. The reason is that only one given element
undergoes G-movement. In the next section, I will look closely at cases where more than
one given element G-moves at the same point in the derivation. Two questions will be at
the center of our discussion: (i) in what order G-movement of more than one element takes
place, and (ii) if a constituent contains more than one given element, can the constituent
move in one step, or do the elements need to undergo separate movements?
3.2 Multiple G-movement
This section looks at cases where more than one given element undergoes G-movement.
We will see that if there is a constituent that dominates more than one given element, it
is preferable to G-move the whole constituent than to G-move each element separately.
Sometimes, however, there is no such constituent that could G-move. I will here investigate
under what conditions more than one given element moves as a single constituent.
Then I will address the question of how given elements move if they must undergo separate
G-movements.
Let’s first look again at example (7), repeated below as (11). The example differs from
the examples that we have seen so far in that it is both the direct object and the indirect
object that must G-move over the verb (recall that the verb must independently move to v,
thus both the objects end up being asymmetrically c-commanded by a new element). The
crucial question is how exactly the two objects, marked by square brackets, move above
the verb.
(11) I have heard that a friend of your cleaning lady stole the new chair from her? Is it
true?
a. Naše uklízečka
[své známé] [tu novou židli] || dala.
our cleaning-woman.Nom REFL friend.Dat that new chair.Acc gave
‘(No, that’s not true.) Our cleaning woman GAVE the chair to her friend.’
The question is whether (i) the two objects move as one constituent, or (ii) whether they
move separately. I will adopt here the former option, arguing that G-movement is as economical
as possible. What I mean by that is that if it is possible to move several given
elements as one constituent, then such movement is preferred to moving them separately.
The condition is stated in (12). With such a condition in place, we need to ask under what
conditions more than one given element may move as one constituent.
(12) Minimize instances of G-movement
4 The derivation continues by merging T and attracting the subject, i.e., the closest element, to Spec,TP.
70

If more than one given element may move as one constituent, then such given
elements must move as one constituent.
Let’s go step by step through the derivation of (11). First, the direct object undergoes short
G-movement over V, as in (13-a). Next, the indirect object is merged and V moves, (13-b).
At this point, both the direct and indirect object need to move. I argue that they move as
one constituent. Thus, it is the whole sister of V – the complement – that undergoes G-
movement at once, as in (13-c). In the next step, V moves to v and the phrase containing
both the direct and indirect object G-moves again. Finally, the subject is merged, (13-d).
(13) a. VP
DO
VP
V
DO
b. VP
V
VP
IO
VP
DO
VP
t V
t DO
c. VP
VP
IO DO t V t DO
VP
V t V P
71

d. vP
subject
vP
VP
vP
IO DO t V t DO
v-V
t V P
VP
VP
t V
t V P
There is, however, a worry that arises about the proposed derivation: If a remnant
phrase may undergo G-movement, why can remnant movement not improve a derivation in
which a given element ends up c-commanded by a new element because of the head movement
constraint? 5 Consider example (16) from section 1.4, repeated below as (14). In this
example, the object ‘that book’ is asymmetrically c-commanded by two given elements.
Further G-movement of the object is, however, blocked by the head movement constraint.
One could imagine a derivation in which the whole VP would move after the direct object
undergoes short G-movement, as in (15). If such movement took place, the given object
would not be asymmetrically c-commanded by any new element anymore. Thus, the partition
between given and new would be perfect. Notice that the constituent that would
undergo G-movement is a complement of the verbal head. Since complements may in general
undergo movement, the derivation would be compatible with our assumptions about
the Czech syntax and about G-movement.
(14) a. What will happen to the book?
b. Marie bude tu knihu dávat Petrovi.
Marie.Nom will the book.Acc give.Inf Petr.Dat
‘Marie will give the book to Peter.’
(15) a. What will happen to the book?
b. #[Tu knihu || dávat Petrovi] bude Marie.
the book.Acc give.Inf Petr.Dat will Marie.Nom
‘Marie will give the book to Peter.’
5 I thank Danny Fox for raising up this question.
72

c. TP
VP
book give to-Peter t book
will
TP
vP
Marie
t V P
There is currently nothing in our system that would rule out the derivation in (15). In chapter
4, I will argue that such movement would not be licensed by the semantic component.
As we will see, G-movement is licensed only if it brings in a semantic interpretation that
would not otherwise be available. It will follow from our semantic system that movement
of VP would not give rise to any semantic interpretation that could not be obtained from
the structure without the VP movement.
Before I turn to structures where several given elements cannot G-move as one constituent,
I want to present two more cases in which G-movement of one constituent dominating
more than one given element is attested. As we have already seen, in Czech, inflected
verbal forms undergo obligatory movement to v, while an infinitive stays in situ. This difference
is reflected in the word order of ditransitive sentences in case only the subject is
new and everything else is given. Consider the examples in (16) and (17). While in the
future tense, the main verb stays inside VP, in the past tense, the main verb evacuates VP.
The result is that if the VP containing both objects undergoes G-movement in the future
tense example, the objects move together with the infinitive. In contrast, in the case of
the past tense, it is only the remnant VP that undergoes G-movement. The corresponding
structures are given in (18) and (19).
(16) Future:
(17) Past:
a. Kdo bude dávat Fíkovi
who.Nom will give.Inf Fík.Dat
‘Who will give Fík presents?’
b. Dávat Fíkovi dárky
give.Inf Fík.Dat presents.Acc
‘Ája will give Fík presents.’
dárky? −→ Inf ≻ IO ≻ DO
presents.Acc
bude
will
|| Ája.
Ája
−→ Inf ≻ IO ≻ DO
a. Kdo dával Fíkovi dárky? −→ IO ≻ DO
who.Nom gave Fík.Dat presents.Acc
‘Who used to give Fík presents.’
b. Fíkovi dárky dávala || Ája. −→ IO ≻ DO
Fík.Dat presents.Acc gave Ája.Nom
‘Ája used to give Fík presents.’
(18) Derivation of (16):
73

TP
VP
TP
give Fík presents
will
vP
Ája
vP
t will
t V P
(19) Derivation of (17):
TP
VP
TP
t V Fík presents
gave
vtP
Ája
vP
t gave
V P
Notice that it is not trivial to characterize movement of a remnant phrase dominating two
given constituents as movement of a given constituent. The VP that moves in (11) is not
straightforwardly given because it is the source of a new verb. As we will see in the following
examples, the only property that characterizes the possible remnant phrase that can
undergo G-movement is that when G-movement takes place, the phrase does not dominate
any new element. Again, we will see in chapter 4 that this will also follow from the way
the semantic component interprets the syntactic output.
We can now turn to the question of under what conditions more than one given element
cannot move as one constituent. I argue that there are two basic types of structures in which
more than one given element cannot undergo G-movement as one constituent. Either (i)
the constituent that dominates only given elements cannot move on independent syntactic
grounds, or (ii) there is no constituent that would dominate only given elements and no new
element.
Let’s first look at the former case, i.e., a case where G-movement is restricted on independent
syntactic grounds. A structure that we can use to test our assumptions is based on
the distribution of VP and vP adverbs. In Czech, certain adverbs can be adjoined both to VP
and vP. Such an adverb is the predicational modifier zase ‘again’. Consider the following
examples.
74

(20) a. Marie otevřela zase dveře. −→ VP adverb
Marie opened again door
‘Marie opened the door again.’
b. Marie zase otevřela dveře. −→ vP adverb
Marie again opened door
‘Again, Marie opened the door.’
There is a semantic difference between the two examples. If again adjoins to VP, as in
(20-a), the utterance presupposes that there was a previous event of opening the door but
it might have been done by someone else than Marie. In contrast, the utterance in (20-b)
presupposes that the door was opened before and it was done by Marie.
The structure that interests us here is a structure in which only the subject is new and
the rest of the clause (the verb, the subject and the adverb) is given. Thus, we want to know
how the derivation proceeds once the new subject is merged and G-movement needs to take
place. Consider the simplified structures in (21). The new subject is marked by a box.
(21) a. [ vP Marie opened [ V P again t V door ]]
b. [ vP Marie again opened [ V P t V door ]]
After the verb moves in order to facilitate G-movement, the relevant structures are as below.
(22) a. [ TP opened [ vP Marie t v [ V P again t V door ]]]
b. [ TP opened [ vP Marie again t v [ V P t V door ]]]
What do we expect next? The question is whether the adverb and the object can move
as one constituent or whether they must move separately. We independently know that
there is no problem with fronting the VP, thus we predict that the whole VP in (22-a)
dominating both the adverb and the object should be able to move as one constituent. In
contrast, movement of an X-bar projection is excluded on independent grounds. Since the
only constituent in (22-b) which contains given elements but no new element is an X-bar
projection, we predict that the given adverb and the given object must move separately. The
expected derivations are schematized in (23).
(23) a. One instance of G-movement:
again object V || subj t V t again t Obj ← ̌[ V P again t V object]
b. Two instances of G-movement:
obj again V || subj t again t V t Obj ← *[ vP ′ again t v [ V P t V obj]
75

As we can see in the examples in (24) and (25), the predictions are borne out.
(24) VP modification:
a. Kdo zavřel znovu dveře? ←− Adv > Obj
who closed again door
‘Who closed the door again?’
b. Znovu dveře zavřela || Maruška. ←− Adv > Obj
again door closed Mary
‘Mary closed the door again.’ [the door was closed before but it wasn’t Mary
who closed it]
(25) vP modification:
a. Kdo znovu zavřel dveře? ←− Adv > Obj
who again closed door
‘Who closed the door again?’
b. Dveře znovu zavřela || Maruška. ←− Obj > Adv
door again closed Mary
‘Again, Mary closed the door.’ [Mary closed the door before]
We can repeat the same exercise with other adverbs, for example with adverbs modifying
telic predicates. An example is given in (26). In English, the adverb almost is ambiguous
between modifying the event of closing the door with the exclusion of the subject and
between modifying the event in which the subject tries to close the door.
(26) Mary almost closed the door.
In Czech, the difference between the two interpretations is overtly marked on the surface by
the position of the adverb with respect to the finite verb, exactly as in the previous examples
with again. The relevant Czech examples are given in (27).
(27) a. Maruška zavřela [ V P skoro dveře].
Mary closed almost door
b. Maruška [ vP skoro zavřela dveře].
Mary almost closed door
The semantic difference is preserved in subject-wh question-answer pairs as well. As we
can see in (28) and (29), the VP attachment results in an answer with a postverbal adverb,
while the vP attachment results in an answer with a clause-initial adverb. Thus, we can
account for the data in the same way as we accounted for the examples with again. While in
(28) both the adverb and the object move as one constituent, in (29), they move separately.
(28) a. Kdo zavřel skoro dveře?
who closed almost door
‘Who almost closed the door (but haven’t finished the closing)?’
b. Skoro dveře zavřela || Maruška.
almost door closed Mary
‘Mary almost closed the door (but she hasn’t finished the closing event be-
76

cause she was interrupted by her mother.).’
c. almost door closed || Marie t V t almost t door
(29) a. Kdo skoro zavřel dveře?
who almost closed door
‘Who almost closed the door (but then decided not to)?’
b. Dveře skoro zavřela || Maruška.
door almost closed Mary
‘Mary almost closed the door (but then she decided not to).’
c. door almost closed || Marie t almost t V t door
Let’s now go step by step through the derivation of the examples in (24-b) and (25-b).
The purpose of the exercise is to see how exactly the system derives the reversed order of
the adverb and the object that we witness in cases with separate G-movements, in contrast
to the preserved word order in case the given elements move as one constituent.
Consider the example in (24-b). First, the VP is merged and the adverb is adjoined
to it. There is no need for G-movement because the structure so far contains only given
elements. In the next step, V moves to v (to check its features) and the subject is merged,
(30-a). At this point, the verb, the adverb and the object all need to move above the new
subject. Only the verb, however, is able to do that because of the head movement constraint
on G-movement, (30-b). After that, the adverb and the object are free to G-move as one
constituent, as in (30-c). The reason why this step is possible is that both given elements
are dominated by a constituent which is able to move. In this case it is VP.
(30) Derivation of (24-b)
a. vP
Marie
vP
closed
VP
again
VP
t V
door
77

. TP
closed
vP
Marie
vP
t v
VP
again
VP
t V
door
c. TP
VP
TP
again t V door
closed
vP
Marie
vP
t v
t V P
We can now proceed with the derivation of the example in (25-b). As we can see in (31),
no G-movement takes place before vP is completed. At this point, the verb, the adverb and
the object need to G-move because they all are asymmetrically c-commanded by the new
subject Mary. Again, the verb must move first, (31-b). After this movement takes place,
the object and the adverb need to move. I have already anticipated that they cannot move as
one constituent. The reason is that the only constituent that dominates the given elements
and no new element is an X-bar projection. Since X-bar projections in Czech cannot move,
the given elements must move separately. The question is whether the given elements move
on nesting or crossing paths. I argue, based on the actual word order we get, that the paths
must be nesting, as in (31-c).
(31) a. vP
Marie
vP
again
vP
v-V
VP
t V
door
78

. TP
T-v-V
vP
Marie
vP
again
vP
t v−V
VP
t V
door
c. TP
door
TP
again
TP
T-v-V
vP
Marie
vP
t again
vP
t v−V
VP
t V
t door
Notice that so far there is nothing in our system that would guarantee the right order of
movement of given elements. The extension condition on G-movement is not on its own
sufficient. In order to account for the word order facts, I propose that if more than one
given element moves in the same point of the derivation, they obey the Path Containment
Condition, as defined in Pesetsky (1982). The definition of the Path Containment Condition
(and the supplementary definitions) are given in (32)–(34).
(32) Path Containment Condition (PCC) (Pesetsky, 1982, ex. (94), p. 309)
If two paths overlap, one must contain the other.
(33) Definition of Paths (Pesetsky, 1982, ex. (69), p. 289)
Suppose t is an empty category locally A-bound by b. Then
79

a. for α the first maximal projection dominating t
b. for β the first maximal projection dominating b
c. the path between t and b is the set of nodes P such that P = {x|(x =α) (x dom.
α & ¬xdom.β)}
(34) Overlapping (Pesetsky, 1982, ex. (93), p. 309)
Two paths overlap iff their intersection is non-null and non-singleton.
In a way, given elements behave as if they were attracted by a feature and obeyed the Attract
Closest Condition. Since we do not have any feature trigger in the system, we must
stay with the descriptive generalization requiring given elements to move on non-crossing
paths. One may wonder whether this is a problem for our system, or whether it is possible
that the source of the path containment behavior is somewhere else than in having a feature
trigger. I do not attempt to answer this question here.
The fact that is crucial for our discussion is that that the Path Containment Condition
results in the reversed word order of the given constituents. In contrast, if given elements
may move as one constituent, their order is preserved. I argue that it is a combination of
these two strategies that captures the variety of word order patterns in Czech.
We can now approach the other case in which given elements cannot move as one
constituent. We will be looking at structures where there is no constituent that would
dominate only given elements. Such an example is given in (35).
(35) a. Co dělal Petr včera s autem?
what did Petr.Nom yesterday with car
‘What did Petr do yesterday with his car?’
b. Petr auto včera || řídil rychle.
Petr.Nom car.Acc yesterday drove quickly
‘Yesterday Petr drove fast his car.’
To see how the derivation proceeds, consider first the corresponding basic word order given
in (36). As we can see schematized in (37), the word order of given elements is partially
reversed (Peter>yesterday>car versus Peter>car>yesterday).
(36) Petr řídil včera rychle auto.
Petr.Nom drove yesterday quickly car.Acc
‘Yesterday Petr drove fast his car.’
(37) Subj-G Obj-G Adv1-G || verb t Adv1 Adv2 t Obj
Let’s now go through the derivation step by step. The derivation is given in (38). We
know from the basic word order that both the adverbs are adjoined to VP (the finite verb,
which is at v, precedes them). The given constituents are marked by boxes. First, the
verb and the object are merged. At this point, the object undergoes G-movement above
the verb, as in (38-a). Then, the new adverb ‘quickly’ is merged and the object needs to
80

move again, (38-b). When the given adverb ‘yesterday’ is merged, no G-movement takes
place. VP is completed and V moves to v, (38-c). At this point, the object and the adverb
‘yesterday’ need to G-move. They cannot move as one constituent, however, because there
is no constituent that would contain only given elements. This is because of the new adverb
quickly. Thus the object and the adverb must move separately. The derivation is completed
by merging the subject, as in (38-d).
(38) a. VP
car
VP
drove
t car
b. VP
car
VP
quickly
VP
t car
VP
drove
t car
c. vP
drove
VP
yesterday
VP
car
VP
quickly
VP
t car
VP
t drove
t car
81

d. vP
Petr
vP
car
vP
yesterday
vP
drove
VP
t yesterday
VP
t car
VP
quickly . . .
We can test on this example a further G-movement property – its strictly cyclic character.
Let’s see what happens if we only minimally change the example in (35-b). Let the subject
be new this time as well. First of all, what do we expect in our system? We are now in the
point of the derivation that corresponds to (38-d), repeated below as (39).
(39) vP
Petr
vP
car
vP
yesterday
vP
drove
VP
t yesterday
VP
t car
VP
quickly . . .
82

Since the subject is new, the derivation cannot stop here. Instead, the verb must move
again, followed by the given object and the given adverb. The interesting question is how
the given elements move. Since there is still no constituent that would dominate only
the given elements, the given elements must move separately. Thus, we predict that their
already reversed order would be reversed again, i.e., the final order would be identical to
their basic word order. This prediction is born out as can be seen in (40). The corresponding
derivation is given in (41).
(40) a. Who and what did yesterday with the car?
b. Včera auto || řídil Petr rychle.
yesterday car drove Petr quickly
‘Petr drove yesterday car quickly.’
(41) a. TP
drove
vP
Petr
vP
car
vP
yesterday
vP
t d.
VP
t y.
VP
t c.
VP
quickly . . .
83

. TP
yesterday
TP
car
TP
drove
vP
Petr
vP
t car
vP
t yesterday . . .
VP
quickly . . .
To sum up, we have seen in this section that G-movement can target a single given
element as well as a constituent dominating more than one given element. I have argued
that movement of a constituent containing more than one given element is preferred to
moving the given elements separately. Whether such a constituent may G-move depends
on two factors: (i) such a constituent must independently be able to move, and (ii) the
constituent may dominate only given element sat the point of G-movement. I have argued
that the variety of word order patterns found in Czech arises from a combination of the two
moving strategies: (i) given elements moving separately, and (ii) given elements moving
as one constituent. Thus, G-movement turned out to be a sufficient tool for deriving the
complex Czech data. But we have also seen that we do not have a sufficient metrics yet
that would decide when exactly G-movement is licensed. In chapter 4 I will show how the
syntax-semantics interface decides when G-movement is licensed and when it is not.
3.3 Summary
In this section, we have seen how G-movement introduces a sort of verb second pattern
in which the given elements are moved around the verbal head, unless such a given element
is the last element merged in the relevant part of the derivation. I have argued that if
more than one given element can move as one constituent, such a derivation is preferred
to a derivation in which the given elements undergo independent movement. In contrast,
if several given elements cannot move as one constituent, they must move independently.
The independent instances of G-movement must obey the Path containment condition. I
argued that a combination of these two strategies – order preserving movement of one con-
84

stituent containing several given element versus order reversing movement of several given
elements – derives the variety of word order patterns found in Czech.
In the next chapter I will address the question of what drives G-movement. I will built
on results from this and the previous chapters, mainly on the conclusion that G-movement
is an extremely local, last resort operation that seems to be insensitive to possible feature
percolation. I will propose that G-movement is free movement that is restricted only by
interface requirements, in particular, by the syntax-semantics interface.
85

Chapter 4
Semantic Interpretation
The proposal to be developed here is similar to recent proposals that attempt to tie information
structure related movement to properties of the syntax-phonology interface (for
example, Cinque 1993; Reinhart 1995; Zubizarreta 1998; Arregui-Urbina 2002; Szendrői
2003). Even though I assume that the phonological facts should be derivable from the
proposed syntactic structure, the syntax-phonology interface is not understood here as the
trigger of overt syntactic movement. I will argue instead that the burden should be put on
the syntax-semantics interface.
The basic idea behind my proposal is that G-movement is free syntactic movement restricted
by semantics in the sense that syntactic items move in order to create a partition
between elements to be interpreted as given and elements to be interpreted as new (cf.
Neeleman and van de Koot 2006 for a similar idea). The partition is interpreted by the semantic
module as a partition between what is presupposed and what is not, while satisfying
the Maximize Presupposition maxim (cf. Heim 1991).
As we will see, however, G-movement is only one part of a system of marking elements
as given. I will concentrate here on coordinations and basic word order structures in which
the same requirement on a partition between given and new applies even though there is no
movement attested.
The chapter is structured as follows. First, I will recapitulate the syntactic results of the
previous chapters and I will highlight the type of data that the current syntactic system cannot
account for. In section 4.2, I will introduce a semantic operator, the purpose of which
is to recursively mark syntactic elements as presupposed. In sections 4.3, I will develop a
system which uses the Maximize Presupposition maxim of Heim (1991) to distribute the
semantic operator and to evaluate syntactic derivations with respect to givenness. Section
4.4 will present the system in more details. Sections 4.5 and 4.6 will formalize the notion
of givenness in Czech and address the question of how givenness in Czech relates to deaccentation
in English. Section 4.7 shortly addresses the role of phonology in the system and
finally section 4.8 concludes the chapter.
86

4.1 Where we stand
In the previous chapters I have introduced G-movement as a tool to derive a hierarchical
partition between given and new material. We have seen that in order to derive the word
order within the given part, movement must be extremely local and cyclic. We have also
seen that this type of movement is a last resort operation in the sense that it takes place only
when it is needed to achieve a distinct semantic interpretation. Crucially, we have also seen
that an element may be given without undergoing G-movement.
I have provided evidence suggesting that G-movement is not feature-driven: G-movement
may take place at any point of the derivation. The only relevant criterion is the information
structure status of a newly (re-)merged item. Interestingly, the movement of α is restricted
by head movement of the head in the projection in which α was originally merged. I
have suggested that we should understand the relation between head movement and G-
movement as a kind of parasitic relation: head movement creates a landing site for α. The
idea is that in case there is no feature attractor, an element must exploit an independently
existing structure.
Since elements can be given without undergoing G-movement, G-movement per se
does not mark an element as given. G-movement only creates a structure in which such an
element may be interpreted as given. The syntactic module does whatever it can do within
its own limitations and it is up to the interface(s) to figure the relevant interpretation and
to license the structure. Notice that, except for the head movement restriction, we have not
seen any constraints on G-movement that are not attested elsewhere in the syntax of Czech.
Is this right, though? Is the only purpose of G-movement really to create a particular
configuration? As we will see, this is indeed so. A piece of direct evidence comes from
coordination facts. Even though there is no movement attested within (and out of) a coordination,
a given element must still linearly precede a new element. Consider the examples
in (2) and (3). In the relevant context, (1), only the verb ‘read’ is given. As we can see, it
must linearly precede the new verb ‘translate’. The examples in (4) are here as a control to
show that if the verb ‘read’ were not trapped within a coordination it would have to move.
Boldface in the following examples stands for given.
(1) Many of my friends have recently decided to change their lifestyle. . .
a. Tak jedna moje kamarádka bude víc číst. ←− context
so one my friend will more read
‘For example, a friend of mine will read more.’
(2) A její přítel bude [číst a překládat]. ←− ̌new > VP & VP
and her friend will read and translate
‘And her boyfriend will read and translate.’
(3) #A její přítel bude [překládat a číst]. ←− # VP & VP
and her friend will translate and read
‘And her boyfriend will translate and read.’
87

(4) a. Číst bude (taky) její přítel .
read will also her friend
‘Her boyfriend will read as well.’
b. #Její přítel bude číst.
her friend will read
The same pattern emerges with coordinated DPs. As we can see from the contrast between
(6-a) and (6-b), a given DP must precede a new DP. Thus, the emerging pattern can be
summarized as in (7), where boldface stands for given.
(5) Na programu byla diskuse o nové učitelce. ←− context
on program was discussion about new teacher
‘The topic of the program was a discussion about a new teacher.’
(6) a. Učitelku a žáky to překvapilo. ←− ̌DP & DP
teacher and students it surprised
‘The teacher and students were surprised by it.’
b. #Žáky a učitelku a to překvapilo. ←− # DP & DP
students and teacher it surprised
‘Students and a teacher were surprised by it.’
(7) Generalization about coordination:
a. A & B
b. #B & A
The facts are, however, more subtle. Recall the discussion about the distribution of pronouns
in section 1.5. We have seen that in general a given DP cannot be asymmetrically
c-commanded by a new element. In such a configuration the DP must be realized as a
pronoun or it must be modified by a demonstrative pronoun. The same holds here. Within
a coordination, a given DP may follow a new element only if the DP is pronominalized, as
in (8).
(8) a. #Žáky a učitelku to překvapilo. ←− # DP & DP
students and teacher it surprised
‘Students and a teacher were surprised by that.’
b. Žáky a tu učitelku to překvapilo. ←− ̌DP & that DP
students and that teacher it surprised
‘Students and that teacher were surprised by that.’
c. Žáky i ji to překvapilo. ←− ̌DP & pronoun
students and her it surprised
‘She (=the teacher) and students were surprised by that.’
The pronominalization facts bring up another important issue. Recall that in section 1.5
we have seen that if a DP coordination is asymmetrically c-commanded by new material, a
given DP is degraded even if it is the first conjunct. Interestingly, coordinated VPs behave
88

differently. As we have seen in (2), there is no problem for a VP conjunct to be asymmetrically
c-commanded by a new material, as long as the given conjunct precedes the new
conjunct.
One might think that the difference between a VP and a DP coordination lies in that
there is no pronominal/anaphoric element that could replace a VP trapped in a coordination.
Before we try to develop a system that could capture the difference between nouns and
verbs, let’s first check whether the existence or non-existence of an anaphoric element is the
relevant difference. If the difference is really between having and not-having an anaphoric
counterpart, we expect that the difference should hold even within the same lexical category
in case the differing lexical items have or do not have an anaphoric counterpart. Such
a category is adverbs. As we can see in the examples in (11), non-pronominalized adverbials
show the same restriction on ordering as DPs and VPs. Interestingly, the restriction
on ordering disappears once we replace the adverbial ‘in Boston’ with pronominal adverb
‘here’, as in (12). 1 Thus, we see the same pattern as with DPs. There is, however, an interesting
contrast. While in the case of a DP coordination, pronominalization is obligatory,
in the case of coordinated adverbs it seems to be optional. Thus adverbials seem to pattern
partially with DPs and partially with VPs. One could argue that the difference might
lie in the uncertainty of judgments related to temporal and spacial indexicals. Even if I
utter the sentence in (11-a) in Boston, it is not clear that being in Boston is the common
ground between the participants of the communication. It might be understood as being
in Cambridge, being at MIT, being in my office etc. In section 4.3 I will argue, though,
that it is possible to account for the difference on independent grammatical grounds. The
difference between having or not having a pronominal counterpart will be only indirect for
the proposal developed here.
(9) My social life in Boston is bearable because. . . ←− context
(10) v Bostonu žije taky moje kamarádka Petra.
in Boston lives also my friend Petra
‘. . . my friend Petra lives in Boston as well.’
(11) a. moje kamarádka Petra žije napůl v Bostonu a napůl v New
my friend Petr lives half in Boston and half in New
‘. . . my friend Petra lives half in Boston and half in New York.’
b. #moje kamarádka Petra žije napůl v New Yorku a napůl v
my friend Petr lives half in New York and half in
‘. . . my friend Petra lives half in New York and half in Boston.’
(12) a. moje kamarádka Petra žije napůl tady a napůl v New
my friend Petr lives half here and half in New
‘. . . my friend Petra lives half here and half in New York.’
b. moje kamarádka Petra žije napůl v New Yorku a napůl
my friend Petr lives half in New York and half
‘. . . my friend Petra lives half half in New York and half here.’
Yorku.
York
Yorku.
York
Bostonu.
Boston
tady.
here
1 In contrast to English, Czech ‘tady’ is an anaphoric element, unlike ‘here’. Non-anaphoric counterpart
of ‘here’ is ‘zde’.
89

To summarize, we have seen that there are two different strategies to mark a syntactic
element in Czech as given. (i) Either an element is in an appropriate configuration (given
precedes new), or (ii) it is lexically marked as given. The appropriate configuration can
be achieved either by base generation or by G-movement. The relevant environments we
have discussed so far are summarized in (13). We have also seen that failing to be in an
appropriate configuration is not necessarily a problem. The facts are more subtle, as we
have seen in the coordination of adverbs.
(13) Summary of environments in which a given element must precede a new element:
a. finite clause (may involve cyclic G-movement)
b. domain defined by verb movement (may involve local G-movement)
c. coordination (no movement)
The syntactic mechanisms introduced so far create the right partition when some G-movement
takes place. It cannot, however, account for the word order constraint in coordination and it
cannot account for the distributional pattern of pronominalized versus non-pronominalized
elements because G-movement is the only mechanism that can improve a suboptimal structure.
The question is whether we can do better.
Let’s first see whether we can do better if we take into account current insights about the
syntax of information structure. There are two main approaches to consider: cartographic
approaches (Rizzi, 1997, 2004; Cinque, 2002; Belletti, 2004; Aboh, 2004, among others)
and interface-driven approaches (Vallduví, 1992; Cinque, 1993; Reinhart, 1995; Erteschik-
Shir, 1997; Zubizarreta, 1998; Arregui-Urbina, 2002; Szendrői, 2003, among others). 2 The
cartographic approaches cannot deal with the Czech word order data. Such an approach
would have problems with ordering within the given part (when should the order be preserved,
when should it be reversed?), with ordering within a coordination (if feature checking
can be done by Agree, why this option is not available for any given element?), and
with the fact that not only given but sometimes also new elements (a finite verb) undergo
movement. The interface-driven approaches (more precisely approaches which rely on the
syntax-phonology interface) deal with the Czech data much better. For example, if we consider
a rather abstract version of the nuclear stress rule, we might be able to account for
most of the ordering patterns. But unfortunately, exactly the cases which cannot be captured
by G-movement, such as ordering within a coordination and the difference between
coordinated DPs and coordinated VPs, would be left aside as well. Furthermore, a phonology
driven approach does not provide any insight for deriving the empirical generalization.
I argue that the missing part in our understanding of givenness is the syntax-semantics
interface. Even though many of the above mentioned authors are concerned with the semantic
interpretation of givenness (or focus), the syntax-semantics interface is not understood
as the key component. Instead, the syntactic component or the phonological component
works in parallel to the semantics. I argue that we indeed do want to put the main burden
2 I do not comment here on optimality theory approaches (Choi, 1999; Downing et al., 2006; Zerbian,
2006, among others) and functionalist approaches (Sgall et al., 1986; Firbas, 1992, among others).
90

on the semantic component, more precisely on the semantics and the pragmatic component.
Once we do this, we do not need to refer at all to information structure in syntax and/or in
phonology.
Notice that we have never defined what it means to be given in Czech. Following
Schwarzschild (1999), I argue that a given element in Czech must have a salient antecedent.
As we will see in section 4.5, however, this is not a sufficient condition for givenness in
Czech. For an element to be given in Czech, the element must be existentially presupposed.
(14) For α to be given,
a. α must have a salient antecedent (cf. Schwarzschild 1999), and
b. α must be presupposed.
Consider now the following English example. The relevant utterance is the utterance
in (15-c). As we can see in (16) the Czech counterpart of (15-c) is infelicitous, unless the
string is reordered as in (16). The crucial question is why (16) is not well-formed.
(15) a. Q: Any news about Mary or her sister?
b. A: Not much. But I’ve heard that. . .
c. the new president fell in love with Mary.
(16) #Nový prezident se zamiloval do Marie.
new president.Nom REFL loved into Marie.Gen
‘The new president fell in love with Marie.’
(17) Do Marii se zamiloval nový prezident.
into Marie.Gen REFL loved new president.Nom
‘The new president fell in love with Mary.’
The answer to this question is in fact far from being trivial. We know that the utterance in
(16) is syntactically well-formed. As the semantic interpretation is concerned, the utterance
is well formed as well. There is no problem with interpreting the utterance irrespectively
of whether or not Marie is presupposed. Furthermore, according to our informal definition
of givenness in (14), if Marie is presupposed, Marie is given. To conclude, with our current
system, there is no reason for Marie to G-move. We predict that the utterance in (16)
should be well formed and felicitous, but this is incorrect. This is true in general: in the
system in place there is no reason for an element to G-move.
I argue that the problem lies in a peculiarity of Czech that is schematically captured in
(18). Roughly, an element cannot be presupposed without structurally higher elements in
the same domain being presupposed as well.
(18) A peculiarity of Czech:
Within a domain [ Dom Y . . . X], if X is presupposed, so is Y.
With respect to the utterance in (16), there are two options that can arise. Either (i) Marie is
presupposed, or (ii) Marie is not presupposed. Let’s evaluate these two options. (i) If Marie
is presupposed, it follows from (18) that ‘new president’ must be presupposed as well. But
91

such an interpretation would lead to Presupposition failure. What about the other option?
If we do not presuppose Marie, nothing goes wrong either in syntax or in semantics. I
argue that in order to exclude option (ii) we need to refer to a pragmatic principle called
Maximize Presupposition, given in (19). I argue that if Marie in (16) is not presupposed
then (16) violates the Maximize Presupposition maxim.
(19) Maximize Presupposition (after Heim (1991))
In context C use the most informative presupposition satisfied in C.
We can thus conclude that (16) is not well formed because Marie cannot be interpreted as
presupposed in this particular syntactic configuration. The only way to interpret Marie as
presupposed is to change the structure, i.e, to move Marie above the new elements.
To formalize the idea about marking givenness introduced in this chapter, we will first
need to derive the descriptive generalization about Czech given in (18). Then I will introduce
a formal evaluation component that decides what structure satisfies the Maximize
presupposition maxim in the relevant context. I will argue for a global comparison system
which will evaluate syntactic structure at the level of a phase.
More concretely, in the next subsection I will derive (18) by introducing a semantic
operator which recursively marks syntactic elements as presupposed. Then I will show
how this operator interacts with Maximize Presupposition. I will also show how the modified
system can account for the Czech cases discussed in chapters 1–3. In section 4.3 I
will show how the modified system can account for the coordination facts that have been a
problem for the original system. Section 4.5 formalizes the notion of givenness in Czech
and in section 4.6 I will address the question of the relation of G-movement in Czech and
deaccentation in English. Finally, in section 4.7 I will show why a syntax-phonology interface
system is not a viable alternative.
It has already been suggested that Maximize Presupposition may license movement
(see Wagner (2005, To appear a) and Wagner (To appear b)). I want to extend the idea to
licensing other grammatical structures as well. Roughly, Maximize Presupposition may be
used for global comparison of different derivations. It is up to the reference set to incorporate
whatever the relevant means of expressing givenness in a particular language are.
The intuition to capture is that there may be more than one grammatical tool to consider
within the comparison set. Thus, while some languages use, for example, morphological
marking (for definite articles) or prosodic tools (deaccenting) as means which can give rise
to a presupposition (can pick up a unique referent from the discourse), other languages may
have other tools. I argue that Czech uses movement (cf. Hlavsa (1975) for a similar idea)
and a linear partition between given and new as such a tool. 3
3 Notice I do not claim that givenness in Czech corresponds to definiteness. Even though there may be a
partial overlap, these are two different notions.
92

4.2 Marking givenness by an operator
Recall that the example in (16) is not well formed no matter whether Marie is presupposed
or not. If Marie is not presupposed the pragmatic principle Maximize Presupposition is
violated. In contrast, if Marie is presupposed, (16) is out because of the peculiarity of
Czech characterized in (18). The question is whether we can derive (18), repeated below
as (20).
(20) A peculiarity of Czech:
Within a domain [ Dom Y . . . X], if X is presupposed, so is Y.
Roughly, we need something that adds a presupposition to an element without affecting
the assertion. I will implement this idea by using a semantic operator that I will call G-
operator.
In principle we could have a semantic operator that could apply anywhere in the structure
and which would mark its sister as given (see Sauerland (2005) for such a proposal for
English). Consider the structures in (21) and (22) which demonstrate such a proposal.
(21) =⇒
given
given
(22)
given
new new . . .
G
given G new new . . .
This structure does not seem to be right because such an operator would not capture
(18) and as a result no movement would be needed. Recall that even though sometimes
elements can be interpreted as given in situ, they usually relocate to the left edge of their
domain.
Furthermore, as the following subsection intends to show the relevant domains for
movement and for spreading presuppositions correspond to a proposition (type < s,t >).
We thus need an operator that can take more than one element (a recursive operator) and
that is sensitive to semantic types in that it terminates on type .
(23)
given
given
G
new
new . . .
93

In this section I will propose such a recursive operator. I will show how the operator
motivates the location of given elements on the left edge of their domain and the fact that
the domains correspond to propositions. In section 4.2.1, I will give arguments for the
domains being propositional. Next, in section 4.2.2, I will define the operator and I will
show how it interacts with the Maximize presupposition maxim.
4.2.1 G-movement domains are propositional domains
So far, we have seen several types of domains within which G-movement takes place. Either
(i) the domain corresponds to a finite clause, which we can take to be a proposition
without any further discussion, or (ii) the domain is whatever is selected by a tense auxiliary,
or (iii) the domain is a small clause. 4
If we assume that a tensed auxiliary selects for a proposition, following Ogihara (1996),
among others, we can conclude that in all the cases discussed so far, the relevant domain
of G-movement is a proposition. The relation is schematized in (24). The examples in
(25)–(27) illustrate the relation between a tense auxiliary and the domain of G-movement.
(24) a. Future:
Aux-v < proposition VP >
b. Past:
Aux-T < proposition vP + VP >
c. Present:
< proposition (CP) + TP + vP + VP >
(25) Future:
(26) Past:
a. What will happen with all the money that was found in the building?
b. Nějaký úředník [ vP bude [ V P peníze || pravidelně posílat opuštěným
some clerk.Nom will money.Acc regularly send lonely
dětem.]]
children.Dat
‘A clerk will regularly send the money to lonely children.’
a. What happened with all the money that was found in the building?
b. Nějaký úředník a já [ TP jsme [ vP peníze || pravidelně posílali
some clerk.Nom and I were money.Acc regularly sent
opuštěným dětem.]]
lonely children.Dat
‘A clerk and regularly sent the money to lonely children.’
(27) Present:
4 I put aside DPs and coordinations where we have not detected G-movement. I will get back to the
coordination facts in section 4.4.
94

a. What happens with all the money that was found in the building?
b. Peníze || posílá pravidelně nějaký úředník opuštěným dětem.
money.Acc sends regularly some clerk.Nom lonely children.Dat
‘A clerk and regularly sends the money to lonely children.’
So far the correlation between auxiliary selection and the domain of G-movement being a
proposition is only suggestive. Even though such a correlation is possible, the relation is
not straightforward. Furthermore, the assumption that tense selects a proposition has been
questioned in recent literature on tense (Kusumoto, 2005).
In the rest of this subsection I will build an additional argument for treating domains of
G-movement as propositions. The argument is based on behavior of propositional modifiers
such as ‘again’. Modifiers such as ‘again’ are known to be able to attach at different
levels of a syntactic structure and their different syntactic position corresponds to different
scope properties. Bale (To appear) has noticed that while the attachment site of modifiers
like ‘again’ may vary, the constituent ‘again’ attaches to is always a proposition. Since only
a proposition gives rise to a presupposition we can test whether a constituent is a proposition
or not by looking at presuppositions the modifier gives rise to.
Thus if the hypothesis about the relation between G-movement and propositions is correct,
we predict that different tenses in Czech should have different presuppositions. Let’s
consider the difference between perfective versus imperfective future tense formation in
Czech. While the imperfect future tense is formed by a v-generated auxiliary and an infinitive,
the perfective future tense is formed by a synthetic finite verbal form. See the
examples in (28).
(28) a. Marie nakoupí (*dvě
Marie.Nom shops.Fut.Perf two
‘Marie will shop.’
b. Marie bude nakupovat dvě
Marie will.3.sg. shop.Inf.Imp two
‘Marie will shop for two hours.’
hodiny).
hours
hodiny.
hours
If we assume that the imperfective auxiliary selects for a proposition and that the subject
is base generated as the specifier of vP, we predict that the structure may give rise to a
presupposition that excludes the subject. In contrast, since in the perfective future tense the
domain of G-movement is bigger than VP, no subject-less proposition is predicted to be
possible. The predictions are schematized in (29). As we will see shortly, the predictions
are borne out.
(29) a. Imperfective Future:
[ vP Subject Aux [ V P again event]]
→ again gives rise to a presupposition in exclusion of the subject
b. Perfective Future:
[ vP Subject verb [ V P again . . . ]]
→ again cannot give rise to a presupposition in exclusion of the subject
95

Consider first English examples with the modifier ‘again’. Certain verbs (according
to Bale, non-stative verbs) allow again to combine with the VP without giving rise to a
presupposition containing the subject. Thus, as in the example in (30), it is enough that
someone hugged Esme to license again in ‘Jon hugged her again.’ For this sentence to be
felicitous there is no requirement on Jon to have hugged Esme before. Thus, the subject
is not part of the relevant presupposition otherwise the sentence with Jon would not be
felicitous (unless there was an event in the past in which Jon hugged Esme).
(30) simplified after Bale’s (28):
a. Jon and his wife love their daughter Esme, and Esme is reassured by overt
expression of their love. For example, yesterday Esme felt reassured when
her mother gave her a hug. The effect was doubled when. . .
b. Jon hugged her again.
In contrast, other verbs (according to Bale, stative verbs) can combine with again at the
VP level only if the relevant presupposition contains also the subject. As we can see in
the example in (31), for someone to love Frank again, the person must have loved Frank
before. The (c) example is here as a control to show that if the subject is not present in the
structure there is nothing wrong with the presupposition triggered by ‘again’.
(31) simplified after Bale’s (47):
a. Seymour’s mother loved Frank, although she was the only one who did. After
a while she no longer cared for Frank. However, Seymour became attached
to the man, and developed strong feelings for him after his mother’s love
subsided. So. . .
b. #Seymour loved Frank again.
c. Frank was loved again.
Czech exhibits the very same cut as we can see in the Czech equivalents of the English
examples, given in (32) and (33).
(32) a. Petr and his wife love their daughter Lucie, and Lucie is reassured by overt
expression of their love. For example, yesterday Lucie felt reassured when
her mother gave her a hug. The effect was doubled when. . .
b. Petr zase objal Lucii.
Petr.Nom again hugged Lucie.Acc
‘Petr hugged Lucie again.’
(33) a. Petr’s mother loved Martin, although she was the only one who did. After a
while she no longer cared for Martin. However, Petr became attached to the
man, and developed strong feelings for him after his mother’s love subsided.
So. . .
b. #Petr
zase miloval
Petr.Nom again loved
‘Petr loved Martin again.’
byl zase
was again
c. Martin
Martin.Nom
Martina.
Martin.Acc
milován.
loved
96

‘Martin was loved again.’
The cut in Czech is identical to the morphologically marked difference between perfective
and imperfective verbs. 5 Thus, while imperfective verbs do obligatory give rise to
a presupposition containing the subject, perfective verbs are compatible with subject-less
presuppositions. Since most Czech verbs have both imperfective and perfective stems we
can test the difference on minimal pairs.
(34) Imperfective version of kick:
a. Marie was kicking Petr for half an hour and then she stopped. Petr got relieved
but then. . .
b. #Jana zase kopala Petra.
Jana.Nom again kicked.Impf Petr.Acc
‘Jana kicked Petr again.’
c. Petr byl zase kopán.
Petr.Nom was again kicked
‘Petr was kicked again.’ 6
(35) Perfective version of kick:
a. Marie kicked Petr once and then she left. Petr got relieved but then. . .
b. Jana zase kopla Petra.
Jana.Nom again kicked.Perf Petr.Acc
‘Jana kicked Petr again.’
We can see the morphological difference more clearly in the future tense because imperfective
verbs form the analytical future in contrast to perfective verbs which form the
synthetic future.
(36) Imperfective version of kick – future:
a. Marie was kicking Petr for half an hour and then she stopped. Petr got relieved
because he didn’t know that. . .
b. #Jana bude zase kopat Petra.
Jana.Nom will again kick.Impf Petr.Acc
‘Jana would kick Petr again.’
(37) Perfective version of kick – future:
a. Marie kicked Petr once and then she left. Petr got relieved because he didn’t
know that. . .
b. Jana zase kopne Petra.
Jana.Nom again will-kick.Perf Petr.Acc
5 I believe that Bale’s conclusion that the relevant partition of verbs with respect to their propositional
properties is not correct. I think that even in English, the relevant difference is between perfective and
imperfective verbs. The difference is, however, not so easily detectable because English does not have any
overt morphological marking that would give a clear cut for Bale’s cases. Bale lists several counterexamples
to his proposal. They all fall under the perfective/imperfective distinction.
6 In fact, the passive form is slightly awkward but this is for independent reason: there is a lexical ambiguity
involving (‘kick’ and ‘dig’) because of which a native speaker prefers an impersonal passive to a regular
passive.
97

‘Jana would kick Petr again.’
To conclude, we have seen that there is a correlation between auxiliary selection and the
size of a propositional domain. I argue that the presuppositional behavior discussed above
provides additional evidence for treating domains of G-movement as propositions.
The other cases where we observed very local G-movement were small clauses and
infinitival complements. Also here, we can safely assume that both small clauses and
infinitival complements are propositions. To sum up, in all the relevant cases, the domain
of G-movement corresponds to a proposition. In other words, given elements are adjacent
to type . This will turn out to be crucial for the way we define the G-operator.
4.2.2 The G-operator
Let’s recapitulate where we stand. We need a meta-language object that would add a presupposition
to a syntactic element. Such an object (i) needs to be able to add a presupposition
to more than one element (in order to capture the pecualiarity of Czech schematized
in (20)), and (ii) it may operate only within a propositional domain (we know that elements
outside of a propositional domain does not need to be presupposed). I propose to
implement such an object as the syncategorematic operator defined in (38). 7
(38) G-operator:
G(B)=
{ λA α : Given(A).G(B A) B is of type < α,β > for some α, β
other than
B
for B of type
The operator takes a constituent B and marks its sister A as Given which rougly means
that it adds a presupposition to A. 8 The operator is defined with respect to a syntactic
constituent. Thus, whatever can be syntactically (and semantically) combined together
qualifies as good arguments for the operator. There is a checking condition involved: the
operator marks A as given only if B is not of type s,t. If B is of type s,t, the meaning of the
G-operator is an identity function which returns B. 9 Notice that it is immaterial whether
the operator is inserted in the narrow syntax or at LF. The only important thing is that it is
syncategorematic in the sense that it is not interpreted compositionally.
Let’s consider a simple derivation to see how the operator works. In the example in (39),
there is only one given element: ‘lollipop’. The corresponding LF structure is given in (40).
I follow here Heim and Kratzer (1998) in treating movement as inducing λ-abstraction.
But I modify their proposal in assuming that the operator is inserted above the λ-abstractor.
7 I am grateful to Irene Heim and Roni Katzir for discussing the properties of the G-operator with me. The
formulation of the operator in (38) is due Roni Katzir.
8 I will introduce a formal definition of Given in section 4.5.
9 To my knowledge, Schwarzschild (1999) was the first to observe that interpreting givenness instead of
focus allows given elements to be interpreted in a recursive fashion.
98

Furthermore, I assume that if there is more than one given element then the G-operator
is above all the λ-abstracts (cf. Nissenbaum (1998), Nissenbaum (2000) and Beck and
Sauerland (2000)). This is necessary for the G-operator not to terminate below the given
elements.
(39) Lízátko našel chlapec.
lollipop found boy
‘A boy found the lollipop.’
(40)
A=lollipop
C
G C’
7 B
boy B’
found t 7
Simplifying significantly, I assume that head movement is not relevant for the interpretation.
I also assume that both lollipop and boy are individuals. The semantic interpretation
is given in (41)–(44). For the sake of the simplicity of the derivation, I treat propositions in
the following derivations as type t.
(41) found = λx.λy. y found x
B’ = λy. y found x 7
B g = 1 iff boy found g(x 7 )
C’ ∈ D et = λx 7 . boy found x 7
(42)
C = G(C’) =G(λx 7 . boy found x 7 ) =
=λh e : Given(h). G([λx 7 . boy found x 7 ](h)) =
=λh e : Given(h). G(boy found h) =
=λh e : Given(h). boy found h
(43) a. A = l (for lollipop)
b. boy = b
where b and l are individuals
(44) AB = 1 iff
C (A) = 1 iff
[ G (λx. b found x 7 )] (l) = 1 iff
[λh e : Given(h). G ([λx 7 . b found x 7 ](h))](l) = 1 iff
[λh e : Given(h). b found h](l) = 1 iff
l is given: when defined = 1 iff and b found l
As we can see, the G-operator marks A (lollipop) as given but then it terminates because
after the lollipop combines with the rest of the clause (boy found), the structure is of type
t, thus the G-operator returns an identity function. Notice that if there was more than one
99

given element, the operator would not terminate. Instead it would mark the next element
as presupposed.
The reader might be puzzled that the operator leaves its complement intact and propagates
upwards. A more conventional idea of a semantic operator is one that applies only to
its complement. Notice, however, that such a binary operator is not uncommon. This is exactly
how, for example, the generalized conjunction (Partee and Rooth, 1983) or ∗-operator
of Beck and Sauerland (2000) work.
The G-operator as it is defined now does two things for us. First of all, it marks everything
from a certain point up as given and it stops at the edge of a propositional domain.
Furthermore, once given elements start moving they need to continue moving. The reason
is that if there were a new element between two given elements it would be marked as presupposed
as well because the operator would not terminate and the new element would be
necessarily marked as presupposed, leading to presupposition failure. Recall that, for the
operator to terminate, the relevant constituent must correspond to a proposition. We have
thus derived the peculiarity of Czech from (20).
Notice, however, that the G-operator per se does not motivate movement. According to
the definition an element can be marked as presupposed even in situ. Consider the following
structure. For the sake of simplicity let’s assume that this particular VP is propositional.
(45) VP s t
B=V set G A=object e
In this particular configuration it is no problem for the G-operator to mark the object as
presupposed while leaving everything else intact.
(46) B = λx e . shoot x
A = movie (individual)
(47) B’ = G(B) =
G(λx e . shoot x) =
λh e : Given(h). G ([λx e . shoot x](h)) =
λh e : Given(h). G (shoot h) =
λh e : Given(h). shoot h
(48) AB = 1 iff
B’ (A) = 1 iff
[λh e : Given(h). does h] (movie) = 1 iff
movie is given: when defined = 1 iff shoot movie
At this point I do not know any principled way to block such a derivation. But the facts
are clear: we want the G-operator to be able to adjoin to an extended projection, not to a
100

lexical head. In a way, we want the operator to behave like a syntactic adjunct. 10
(49) A G-operator cannot adjoin between a lexical head and its complement.
The ban on adjoining the G-operator between a lexical head and its complement predicts
that in the case of basic word order, anything except for the most embedded element can be
given. This follows from the fact that there is no way to insert a G-operator so that it could
scope over the most embedded element. This is a welcome prediction because an utterance
in which everything is given is perceived as infelicitous. 11
The question that arises is how the grammar knows where to introduce the G-operator.
I argue that the operator’s distribution is determined by Maximize presupposition, repeated
below.
(50) Maximize Presupposition (after Heim (1991))
In context C use the most informative presupposition satisfied in C.
The consequence of the maxim is that the presuppositions we consider are scalar presuppositions.
12 The reason is that the maxim requires the speaker to use the logical form which
marks the strongest presupposition compatible with the common ground (Stalnaker, 1973,
1974). The question is how we restrict the set of presuppositions that are relevant for a
particular utterance. Recent proposals based on Maximize Presupposition, for example,
Sauerland (2003) and Percus (2006) build the relevant set by replacing a lexical item with
its scalar mates (Horn, 1972) within a fixed structure.
Replacing scalar mates within a fixed syntactic structure would not, however, work in
our case. I argue that if we want the maxim to regulate the distribution of the G-operator, the
reference set must contain different syntactic (syntactic in the broad sense) derivations (see
Reinhart (1995); Fox (1995); Reinhart (2006) for proposals using reference set computation
over structures). 13 Thus, we need a definition of a reference set that would allow us to
compare different structures with and without a G-operator. Such a definition is given in
(51).
10 This seems to be a reasonable assumption. While some languages might be able to adjoin their semantic
operators to any syntactic element, others might be more restrictive. See, for example, Büring and Hartmann
(2001), for arguments that while in English focus sensitive particles may be adjoined to any syntactic element,
in German they can be adjoined only to a maximal verbal projection.
11 Notice that we have just derived the fact that given elements do not bear the main sentential stress which
in Czech falls on the rightmost prosodic word, which in turn corresponds to the most embedded part of a tree.
I will comment on prosody versus givenness more in section 4.7. Notice also that in proposals based on focus
projection or in cartographic approaches, the fact that anything except for the most embedded part of the tree
can be given cannot be derived.
12 I treat the Maximize presupposition maxim as a primitive but see Schlenker (2006) for arguments that
it might be possible to derive the maxim from neo-Gracean reasoning, if we take into an account the notion
of common belief of Stalnaker (2002). But as even Schlenker admits, the results are inconclusive. See
also Magri (2007) for further arguments that Maximize Presupposition cannot be reduced to neo-Gracean
reasoning.
13 I assume that both Sauerland (2005) and Wagner (To appear) define their reference set as a set of derivations
as well. Neither of them is explicit about this, though.
101

(51) Reference set for Maximize Presupposition evaluation
For the purposes of Maximize Presupposition, the reference set, toward which
Maximize presupposition is evaluated, consists of all derivations
a. that are based on the same numeration and free insertion of a G-operator, and
b. that make the same assertion.
I assume that the G-operator is not part of the numeration but it is a syncategoramatic operator
which the semantic module can introduce without violating inclusiveness (for example,
Chomsky (2000)). A crucial part of the proposal is that the semantic module has the capacity
to license an otherwise illicit structure but only if there is no other way to achieve the
desired interpretation (see also Fox (2000)). In our case, the illicit operation under discussion
is G-movement. We have derived the fact that G-movement is allowed only if it affects
the semantic interpretation. Now we can define the relevant condition more precisely.
(52) Economy condition on G-movement:
The only structure that is allowed is the structure that has the smallest number of
G-movements and leads to the relevant interpretation (i.e., assertion and presupposition).
The new condition on G-movement has several welcome consequences. First of all, we
no longer need to distinguish between movement of a head for givenness and movement
of a head which facilitates G-movement. Under the current definition, the only thing that
matters is whether the resulting structure allows insertion of a G-operator in a position that
would not be available otherwise. Whether the movement affects a new or a given element
is irrelevant. To see this, consider the following example and its derivation.
(53) a. What about Petr?
b. Petra || vítá Marie.
Petr.Acc welcomes Marie.Nom
‘Marie welcomes Petr.’
Assuming that a T projection does not need to be inserted in the present tense, 14 the derivation
proceeds as in (54). First, the given object moves over the verb, as in (54-a). Then
the verb moves to v. Now the object needs to move again, (54-b). Notice that it is not
possible to mark the object by a G-operator within VP: since VP is not of an atomic type,
the operator would necessarily affect structurally higher material (the verb and the subject;
the elements in the scope of a G-operator are marked by a box). After the object moves to
vP, the subject is merged, (54-c). In principle the derivation might be able to stop here. The
problem is that, in this configuration, there is no position into which a G-operator could be
inserted without marking the subject as given as well (which would lead to Presupposition
failure). Thus, there is no choice other than to continue the derivation: the verb moves to
T, the object moves to Spec,TP and a G-operator can finally be inserted in a position which
satisfies Maximize presupposition without leading to Presupposition failure, (54-d).
14 I assume a grammar in which a functional projection is projected only if it is associated with overt
material or if it is selected by a higher head. Cf. for example Wurmbrand (2006, To appear).
102

(54) a. VP
Petr
VP
V
t Petr
b. vP
Petr
v-V
vP
t Petr
VP
*G
VP
t V t Petr
c. vP
Subject
Petr
vP
v-V
vP
t Petr
*G
VP
VP
t V t Petr
103

d. TP
Petr
G
v-V
TP
Subject
vP
t Petr
vP
v-V
vP
VP
t Petr
t V
̌G
VP
t Petr
Thus, we can strengthen our previous claim in that there is no direct relation between being
given or new and undergoing G-movement. Syntactic G-movement is free movement. It is
only up to the semantic module to decide whether such movement is licensed or not.
We are now in a position to understand why certain derivations cannot be improved by
moving a phrase containing both given and new elements. Consider again example (14)
from chapter 3, repeated below as (55).
(55) a. What will happen to the book?
b. Marie bude tu knihu dávat Petrovi.
Marie.Nom will the book.Acc give.Inf Petr.Dat
‘Marie will give the book to Peter.’
The question is why the whole VP cannot move as in (56). In such a derivation, no new
element would asymmetrically c-command ‘book’. Thus, the partition between given and
new would be perfect.
(56) a. What will happen to the book?
b. #[Tu knihu || dávat Petrovi] bude Marie.
the book.Acc give.Inf Petr.Dat will Marie.Nom
‘Marie will give the book to Peter.’
104

c. TP
VP
book give to-Peter t book
will
TP
vP
Marie
t V P
But we already know that the perfect partition is irrelevant here. The reason is that the
only place where the G-operator may be inserted is between ‘book’ and ‘give’, as in (57).
Any other position would lead to Presupposition failure. The position is, however, already
available after the object G-moves within VP. Moving the whole DP does not bring in any
interpretation that would not already be available after the first instance of G-movement.
Therefore, G-movement of the whole VP is not licensed.
(57) TP
VP
book G give to-Peter t book
will
TP
vP
Marie
t V P
The same reasoning accounts also for the assumption that a constituent containing several
given elements may G-move only if it contains only given elements. If there were any new
element, a G-operator would have to be inserted within the moving constituent, i.e., in the
position in which it would have been inserted if the whole constituent did not undergo G-
movement at all.
Let’s summarize where we stand. We have an operator that can motivate given elements
to be located in the left edge of its propositional domain and we know informally how the
distribution of the operator can be regulated by the Maximize presupposition maxim. The
open question is what happens if there is more than one propositional domain per finite
clause. The prediction is clear. In principle, any propositional domain might have its own
G-operator in the same way as it can have an independent linear partition between given
and new. Consider the example in (58).
(58) Marie bude knihy prodávat
Marie will books sell
‘Marie will sell the books.’
.
105

The example in (58) is felicitous in the following two contexts:
(59) Context I: only books given
What will happen to the books?
(60) Context II: everything given except for sell
What will Marie do with the books?
Thus, there is one syntactic structure that can be interpreted in two different ways. The
interpretations differ only in the number of G-operators in the structure. Either (i) there is
only one operator, terminating at VP, or (ii) there are two different G-operators – one per
propositional domain. 15
(61) Context I: One G attached to the main spine:
TP
Mary
T
TP
will
vP
books
VP
G
terminating point
VP
VP
sell t books
(62) Context II: Two Gs attached to the main spine:
terminating points
TP
Mary
TP
G
TP
T
vP
will
VP
books
VP
G
VP
sell
t books
15 I assume that ‘will’ is not marked for givenness, otherwise the structure would have to contain three
G-operators.
106

In the same way that there can be two separate G-operators, there can be two domains
of G-movement. Thus, for example, there is no problem with having separate G-operators
and separate G-movement in a matrix and in an embedded clause, as can be seen in (63).
In this sentence, the given elements Petr and Marie are objects in two different clauses and
they G-move within their clause, resulting in two independent partitions between given and
new.
(63) a. Do you know anything about Petr and Marie?
b. Náhodou jsem slyšel, že Petrovi
accidentally Aux.1sg heard that Petr.Dat
Marii || zaměstnali v ABB.
Marie.Acc employed.1pl in ABB
|| říkala
told
nějaká
some
paní,
lady
že
that
‘I accidentally heard that some lady told Petr that Marie got employed in the
ABB.’
Similarly, in infinitival structures, some of the structurally higher elements can be given as
well, or they can undergo G-movement, as in (64).
(64) a. Do you know what Mary did with her famous boat?
b. Marie || se pokusila lod’ || prodat.
Marie REFL tried boat to-sell
‘Marie tried to sell the boat (but no one wanted to buy it).’
c. Marii || nařídil soud lod’ || prodat.
Marie.Dat ordered court.Nom boat.Acc sell
‘(You won’t believe it but) a court ordered Marie to sell the boat.’
We can iterate the insertion of the G-operator even further, as in (65). In this case, one
operator must be adjoined within the DP coordination (only one conjunct is given, not the
whole coordination) and another operator is adjoined within the VP (below boat).
(65) a. Do you know what Mary is planning to do with her broken boat?
b. Marie || a nějaký automechanik budou lod’ || v létě opravovat.
Marie and some car mechanic will boat in summer repair
‘Marie and some car mechanic will repair the boat in the summer.’
To summarize, we now have an informal way to characterize cases in which G-movement
is licensed and we have a semantic operator which allows us to mark more than one G-
element as presupposed. In the next section, I will develop a formal system which allows
us to compare structures with respect to Maximize presupposition in a more precise way.
Before we get to the next section notice that we have derived the size of a domain of
G-movement independently of head movement. The question is whether we still need the
restriction on head movement or not. Also, it is not clear whether we can independently
derive the fact that G-movement is very local.
It looks like the head movement restriction only doubles the locality restriction imposed
by the G-operator. Is this really true? Recall that the domain of G-movement varies in the
107

past tense depending on the presence or absence of the auxiliary. The relevant examples
are repeated below.
(66) 3sg.:
(67) 1pl.:
a. What happened to the boat that got damaged in the last storm?
b. Lod’ opravil jeden technik.
boat.Acc repaired one technician.Nom
‘A technician repaired the boat.’
a. What happened to the boat that got damaged in the last storm?
b. Jeden technik a já jsme lod’ opravili .
one technician.Nom and I Aux.1pl boat.Acc repaired
‘A technician and I repaired the boat.’
I assume that example (66) is a vP, while the example in (67) corresponds to a TP. Even
though this is a non-standard assumption it follows from the hypothesis that functional
projections are merged only when needed. Thus it looks like that we could reduce the restrictions
on G-movement to properties of the G-operator. But I hesitate to take this step.
The reason is that, at this point, I do not have a good understanding of free syntactic movement.
It may well be possible that for a language to have free syntactic movement, the
language must also have flexible head movement. I thus leave the answer for future research.
Another issue is whether we still need G-movement to be very local. I argue that this
requirement is needed in order to derive word orders in which a given element is externally
merged immediately above an internally merged given element. Such an example is
repeated below.
(68) a. How did the boy get the lollipop?
b. Chlapec lízátko
boy.Nom lollipop.Acc
|| našel.
found
‘The boy found the lollipop.’
(69) a. [ vP object verb [ V P t V t obj ] ]
b. [ vP subject object verb [ V P t V t obj ] ]
One might argue that, with respect to the G-operator, it is immaterial whether the object
moves above the verb before or after the subject is merged. This is not true, however. The
difference lies in the position of the λ-abstract. If the object moves over the subject, the
λ-abstract would have to be inserted between the subject and the object. As a result, the
G-operator would terminate before it reached the object. Consider the structure in (70).
108

(70) st
terminating point
object
e,st
λ
st
subject
G . . .
Thus if there is only one G-operator per propositional domain (which seems to be correct
considering the nature of the G-operator, though this is hard to test), unless the object
moves before the subject is merged the object could not be interpreted as presupposed. I
therefore conclude that we need an independent requirement which prefers derivations with
very local movement to derivations with a non-local G-movement.
Another issue that has remained unanswered so far is when exactly the Maximize presupposition
evaluation takes place. I assume that the evaluation may happen at the end of
a phase. This allows limitation of the global comparison only to structures which can be
derived within a phase from the same numeration. I assume this is possible under the assumption
that phases correspond to propositions. But nothing in our system depends on this
choice. If it turns out that presuppositions are computed only at the level of the utterance,
the system can be straightforwardly adjusted to that.
4.3 The evaluation component
In the previous section we have concentrated on defining a given operator and on introducing
such an operator within the logical form. In this section, I will set up a formal
evaluation component for the purposes of Maximize presupposition. This step will complete
the modification of the originally purely syntactic system. In the rest of this section I
will examine whether the modified system can account for the data from chapters 2 and 3
that we originally covered by G-movement.
To see exactly how the evaluation component works we need not only to define the reference
set but also to have a clear metrics that we can use to compare different derivations.
We already have a definition of a reference set. The definition is repeated below.
(71) Reference set for Maximize Presupposition evaluation
For purposes of Maximize Presupposition, the reference set, toward which Maximize
presupposition is evaluated, consists of all derivations
a. that are based on the same numeration and free insertion of a G-operator, and
b. that make the same assertion.
In order to have a clear metrics I will define two constraints with respect to which each
candidate within a reference set will be evaluated. 16 First, we need to guarantee that all
16 The reader should not be misled by the terminology. The system developed here is not an optimality
109

derivations are syntactically well-formed. Such a condition is not strictly speaking a part of
the evaluation component, but it is a necessary precondition for a derivation to be considered
for evaluation. The reason why the constraint belongs here, however, is that licensing
of G-movement happens only at the interface.
(72) SYNTAX:
the reference set may contain only grammatically well-formed structures
Then we need to guarantee that the relevant presupposition would be maximized. This
requirement consists of the interaction of two separate requirements: (i) mark everything
given as presupposed (make sure that each given element would give rise to the desired
presupposition), and (ii) avoid marking new elements as presupposed (avoid Presupposition
failure). I will call this constraint INTERPRETATION.
(73) INTERPRETATION:
a. a given element must be marked as presupposed (either lexically or by a G-
operator) [≈ Maximize Presupposition]
b. a new element cannot be marked as presupposed [≈ Presupposition Failure]
We are now in a position to evaluate the system in place and see whether it can account
for the facts we have considered so far. Let’s start with a simple case of a ditransitive
construction in which everything is given except for the indirect object. Since in Czech
a direct object follows an indirect object, it is enough if the direct object moves above
the indirect object. The desired interpretation is given in (74). The boxes correspond
to elements that we want to be interpreted as given; the structure is presented in the basic
Czech word order. The relevant reference set is given in (75). I list here and in the following
examples only a few candidates that are most relevant for the evaluation. The first candidate
is a derivation in which no G-movement takes place and no G-operator is inserted, (75-a).
The second candidate is a derivation with a G-operator inserted above the indirect object but
with no G-movement, (75-b). The third candidate is a derivation with local G-movement of
the indirect object and a G-operator adjoined immediately below the moved object, (75-c).
The last candidate differs from the previous candidate in that G-movement is not local, but
instead the movement is cyclic and crosses several given elements, (75-d).
(74) Desired interpretation:
Peter gave to-Mary book .
‘Peter gave the book to Mary.’
(75) Reference set of (74):
a. No G-operator and no movement:
Peter gave to-Mary book
b. G-operator and no movement:
Peter gave G to-Mary book
theoretic system. I am using the notion of constraints as a technical tool for explicit evaluation of derivations.
In the end of the day we will see that for a structure to be felicitous in the relevant context no constraint may
be violated.
110

c. G-operator and local G-movement:
Peter gave book G to-Mary
d. G-operator and non-local G-movement:
book gave Peter G to-Mary
The evaluation of the reference set is schematized in (76). As we can see, the derivation
without movement or any G-operator is syntactically well formed, but it fails with respect
to the Maximize presupposition maxim (nothing is marked as presupposed). Similarly, the
second candidate fails with respect to the maxim. Even though, this time, two given elements
are correctly marked by the operator, there is one given element – the direct object –
which remains unmarked. The third candidate, on the other hand, satisfies both the syntactic
requirements as well as the pragmatic requirements. All given elements are marked and
the syntax is well formed. The last candidate does well with respect to the interpretation
but the syntax is not well-formed. The reason is that the derivation contains two instances
of G-movement that cannot be licensed. As we can see by comparison with the candidate
in (75-c), the desired interpretation is obtained without any additional movement. Since
G-movement does not come for free, the candidate in (75-d) loses in comparison with the
candidate in (75-c).
(76) Evaluation of the reference set of (74):
Reference set SYNTAX INTERPRETATION
a. ̌ fails
b. ̌ fails
c. ̌ ̌
d. fails ̌
Let’s now consider a minimally different structure in which it is only the direct object
which is given. All other elements are new in the discourse. The relevant reference set
is given in (78). The candidates are the same as in the previous case, (74). There is a
candidate without G-movement and without any G-operator, (78-a), a candidate without
G-movement and with a G-operator, (78-b), a candidate with local G-movement and a G-
operator, (78-c), and finally a candidate with cyclic G-movement and a G-operator, (78-d).
Again, only a few relevant candidates are considered.
(77) Desired interpretation:
Peter gave to-Mary book .
‘Peter gave the book to Mary.’
(78) Reference set of (77):
a. No G-operator and no movement:
Peter gave to-Mary book
b. G-operator and no movement:
111

Peter
gave G to-Mary book
c. G-operator and local G-movement:
Peter gave book G to-Mary
d. G-operator and non-local G-movement:
book G gave Peter to-Mary
The evaluation of the reference set in (78) is schematized in (79).
(79) Evaluation of the reference set of (77):
Reference set SYNTAX INTERPRETATION
a. ̌ fails
b. ̌ fails
c. ̌ fails
d. ̌ ̌
The candidates in (78-a) and (78-b) pattern with the candidates in (75-a) and (75-b). Thus
they fail because the direct object is not marked as given. The candidate in (78-c) is different:
it is syntactically well-formed and the direct object is marked as presupposed. The
problem with this candidate is that the G-operator marks not only the direct object as presupposed
but also the structurally higher new elements. The reason is that there is no early
terminating point for the operator (the structure is in the past tense and there is no auxiliary).
Thus, the object is forced to move above all the new elements. The cyclic G-movement is
licensed in this case because the relevant interpretation is not available otherwise.
Notice that while in the previous discussion we needed to refer to the presence of a new
element asymmetrically c-commanding the given element, no such constraint is needed
anymore. The locality of the movement is determined by terminating properties of the G-
operator. To appreciate this point, consider the following case where only an object is given
but in the same time, the object is blocked within an infinitive. The relevant reference set
is given in (81).
(80) Desired interpretation:
Mary managed to-burn chair
‘Mary managed to burn the chair.’
(81) Reference set of (80):
a. No G-operator and no movement:
Mary managed to-burn chair
b. G-operator and no movement:
Mary managed to-burn G chair
c. G-operator and local G-movement:
112

Mary managed chair G to-burn
d. G-operator and non-local G-movement:
chair G managed Mary to-burn
The reference set contains structures parallel to the reference sets for (74) and (77). Thus,
there is again a candidate without G-movement and without any G-operator, (81-a), a candidate
with a G-operator and no G-movement, (81-b), a candidate with local G-movement
and a G-operator, (81-c), and finally a candidate with cyclic G-movement and a G-operator,
(81-d).
(82) Evaluation of the reference set of (80):
Reference set SYNTAX INTERPRETATION
a. ̌ fails
b. ̌ fails
c. ̌ ̌
d. fails ̌
As in the previous cases, the derivations without G-movement are not felicitous. The crucial
pair of candidates to compare is the structure in (81-c) and the structure in (81-d). Even
though the object is the only given element – thus this example patterns with (77) – , for
the object to be interpreted as given it is enough to undergo only local G-movement. The
difference between the derivation of (77) and the derivation of (80) is in differing propositional
domains at which the G-operator can terminate. In case the object is trapped within
the infinitival VP, as in (80), the relevant propositional domain is the infinitival VP. Thus,
local G-movement to the edge of the VP is enough for obtaining the correct interpretation.
So far so good. But there is an alternative I would like to rule out, namely having the
G-operator as the trigger of syntactic movement. If this were the case, for α to be given, α
would always have to move to G. We have already seen examples of structures in which G-
movement is blocked, such as coordination, which argued against treating the G-operator
as a syntactic trigger in a Probe-goal sense. Other relevant structures to consider are structures
in which no G-movement is required, i.e., basic word order structures.
We have already seen in section 1.2 that in basic word order the partition between given
and new can fall, in principle, at any position. Consider the example in (83) which lists the
possible positions of the partition within a ditransitive clause with a temporal adverb.
(83) a. Maruška včera dala dětem knihu.
Maruška yesterday gave children book
‘Yesterday Mary gave children a book.’
b. Maruška || včera dala dětem knihu.
c. Maruška včera || dala dětem knihu.
113

d. Maruška včera dala || dětem knihu.
e. Maruška včera dala dětem || knihu.
As we can see, there are four possible partitions available. I argue that the partition
positions correspond to positions in which the G-operator may be inserted. Thus, we now
have a formal tool to understand the basic word order pattern. I argue that the exact position
of the G-operator (or of several G-operators in case there is a terminating point somewhere
in the structure) depends on what is presupposed. In other words, the insertion of a G-
operator is regulated by the Maximize Presupposition maxim. If the wrong position is
selected, either some given element will not be marked as presupposed or a new element
will be incorrectly marked as presupposed.
To see this in more detail, let’s consider a scenario in which the subject and the adverb
are given, the rest is new (Tell me what Mary did yesterday.), (83-c). The relevant reference
set is included in the evaluation table in (84).
(84) Evaluation of the reference set of (83):
Reference set SYNTAX INTERPRETATION
Mary G yesterday gave children book ̌ fails
Mary yesterday G gave children book ̌ ̌
Mary yesterday gave G children book ̌ fails
Mary yesterday gave children G book ̌ fails
Since there is no G-movement, all candidates are syntactically well-formed. There is, however,
only one candidate which does well with respect to the Maximize presupposition
maxim. The felicitous candidate is the candidate with a G-operator adjoined immediately
below the adverb. Only this position guarantees that both the subject and the adverb will
be presupposed and the rest of the clause will be understood as new. Notice that setting
up the evaluation with respect to the Maximize Presupposition maxim is crucial here. For
the structure to be correctly interpreted it is crucial to know what presupposition is to be
matched. There is nothing in the syntactic module that would have direct relevance for the
position of the G-operator.
To conclude, we have seen in detail the evaluation of several cases involving G-movement
which were set up to test the hypothesis that to choose the right derivation it is enough to
refer to the G-operator and to the Maximize presupposition maxim. As far as syntax is
concerned, any well-formed derivation is possible as far as it is licensed by semantics (and
pragmatics). Thus by modifying the syntactic system we do not lose anything. The question
is whether we gained something in empirical coverage.
4.4 The G-operator and coordination
In the previous section we have seen that the modified system copes well with the G-
movement data.The question is whether the system can account for coordination facts as
114

well. Recall that there are two relevant restrictions to consider. (i) A given XP must precede
a new XP in a coordination. (ii) A given VP can be trapped in a coordination even if it is
asymmetrically c-commanded by new material. In contrast, a given DP is degraded in
such a configuration and must be pronominalized. The crucial examples are repeated in
(85)–(88). Given elements are in bold font. The structure of the data is schematized in
(89).
(85) Mary had some unpleasant experience with the new teacher and she did not want
to go to the meeting only with her. In order to avoid any conflict,. . .
a. #Pozvala učitelku a ředitelku. ←− # new > DP & DP
invited.3sg.F teacher.Acc and director.Acc
‘She invited the teacher and a director.’
b. Pozvala ji a ředitelku.
invited.3sg.F her.Acc and director.Acc
‘She invited her and a director.’
c. Pozvala tu učitelku a ředitelku.
invited.3sg.F that teacher.Acc and director.Acc
‘She invited that teacher and a director.’
(86) A friend of mine has decided to change her lifestyle. She will read. . . ←− Context
(87) A její přítel bude číst a překládat.←− new > VP & VP
and her friend will read and translate
‘And her boyfriend will read and translate.’
(88) #A její přítel bude překládat a číst.←− # VP & VP
and her friend will translate and read
‘And her boyfriend will translate and read.’
(89) a. #
α New
&P
DP
&P
&
DP
b.
α New
&P
VP
&P
&
VP
115

c. # &P
XP
&P
&
XP
Let’s start with the case in (89-c), i.e., the restriction that a given conjunct must precede
a new conjunct. The crucial point is that for an XP to be marked as presupposed, it must
be an argument of a G-operator. The operator can be attached either (i) immediately above
the given XP, or (ii) it can be attached to the whole coordination, or (iii) it can be attached
below the first conjunct. The corresponding structures are given in (90). The elements in
the scope of the operator are marked by boxes.
(90) G-attachment within a coordination:
a. immediately above the given XP:
&P
Conj1
&P
& G Conj2
b. to the main clausal spine (to the whole coordination):
&P
G
Conj1
&P
&
Conj2
c. immediately below the first conjunct:
&P
Conj1
G
&P
&
Conj2
Let’s consider each case in turn. If the G-operator attaches above the second conjunct, as
in (90-a), the SYNTAX constraint is satisfied but the INTERPRETATION constraint is violated
because the given conjunct is not marked as presupposed while the new part of the
coordination is marked as presupposed. Similarly, in (90-b), too many elements are marked
as presupposed resulting in Presupposition failure. Attaching the G-operator immediately
below the first conjunct does not help either, (90-c).
116

There is simply no way the G-operator could be inserted and the correct interpretation
be obtained without violating either SYNTAX or INTERPRETATION. The only structure
that satisfies both SYNTAX and INTERPREATION is the structure in which the given conjunct
precedes the new conjunct and the G-operator is inserted immediately below the first
conjunct, as in (91).
(91) &P
Conj1
G
&P
&
Conj2
The evaluation of these possible structures is schematized in the table in (92).
(92) Evaluation of [XP and XP]
Reference set Syntax Interpretation
(90-a) ̌ fails
(90-b) ̌ fails
(90-c) ̌ fails
(91) ̌ ̌
As we can see, the G-operator and the Maximize presupposition maxim can take care of
the fact that a given conjunct must precede a new conjunct.
So far so good. Now we have an account of a subset of the restrictions on coordination,
namely, the word order restriction within a coordination. The remaining question is
why there is a difference between a DP and a VP coordination and why adverbs pattern
partially with DPs and partially with VPs. We know that there is a way to mark a DP conjunct
as presupposed in case there is no new material asymmetrically c-commanding the
coordination. Thus, the problem does not lie in the marking itself but must be related to
the presence of new material. We know that the presence of structurally higher new material
does not matter in case a DP undergoes local G-movement. But in such a case, the
DP is on the edge of a propositional domain. Based on this fact, I suggest that the difference
between DP coordination and VP coordination is a difference in their semantic types.
Since a DP coordination is not a propositional domain, the G-operator may escape from
the coordination and mark the structurally higher new material as given as well. Thus, in
case a DP coordination is not on the left edge of a propositional domain, marking the first
conjunct with a G-operator is impossible because the scope of the operator would always
affect a bigger structure, resulting in an infelicitous interpretation. The only way to avoid
the problem of marking new material with the G-operator is to replace the given DP with
a pronoun. Recall that pronouns do not need to be marked by the operator in order to be
given. It is their lexical entry which gives rise to the required presupposition.
117

There is, however, one problem with the system in place. It is not clear how we can
guarantee that the operator within a DP coordination does not mark the predicate the coordination
combines with as presupposed. Everything depends on the semantic type we
assume for a DP coordination. If we assume that a DP coordination is a generalized quantifier,
then the G-operator would necessarily over-generate. On the other hand, if we assume
that such a coordination is of type e, the operator cannot apply to the first conjunct because
it is not defined for atomic semantic types. I suggest that the problem is avoided if we treat
a DP coordination as a boolean coordination, i.e., a sum of individuals. I leave, however,
the actual implementation for future research.
In contrast to a DP coordination, a verbal coordination may be propositional. If this is
the case, then there is no problem with adjoining the G-operator below the first conjunct.
But can we safely assume that all relevant VP coordinations correspond to a proposition?
Let’s consider again the example (1), repeated below as (94).
(93) Many of my friends have recently decided to change their lifestyle. . .
a. Tak jedna moje kamarádka bude víc číst. ←− context
so one my friend will more read
‘For example, a friend of mine will read more.’
(94) A její přítel bude číst a překládat.
and her friend will read and translate
‘And her boyfriend will read and translate.’
This case is straightforward because we know independently that the future tense auxiliary
bude ‘will’ selects for a proposition. 17 The question is whether the same is true for other
types of verbal coordinations. The most suspicious case is a coordination which is not
selected by a tense auxiliary.
(95) Many of my friends changed their lifestyle several years ago. . .
a. Tak jedna moje kamarádka víc četla. ←− context
so one my friend more read
‘For example, a friend of mine read more.’
(96) A její přítel četl a překládal.
and her friend read and translated
‘And her boyfriend read and translated.’
Even though the examples in (94) and (96) look parallel, they are not. There is a difference
in their interpretation. The coordination of the future infinitives does not impose any
requirement on the implicit object of reading and translating. In contrast, the past tense coordination
requires the implicit objects to be identical. We can see the contrast on a possible
continuation of the discourse in (94), given in (97), and the continuation of the discourse in
(96), given in (98). While it is felicitous to say that he will read something different than
he will translate, it is odd to say that he read something different than he translated.
17 I put aside the option that the coordination is a coordination of something bigger than VP, i.e., a structure
which originally contained the auxiliary.
118

(97) a. And what will he read and translate?
b. Bude číst detektivky a překládat romány.
will-he read detective mysteries and translate novels
‘He will read detective mysteries and translate novels.’
(98) a. And what did he read and translate?
b. #Četl detektivky a překládal romány.
read detective mysteries and translated novels
‘He read detective mysteries and translated novels.’
I do not know at this point how to account for the difference between past and future.
Intuitively, the past form of ‘read’ is not given in the same sense as the future form of ‘read’.
There is something about the shared object that seems to make a difference for givenness.
My hope is that these cases contain coordination reduction and the contrast between implicit
objects in past and in future is related to a structural difference. I leave the question
of differences among VP coordinations as a puzzle for future research.
For now, let’s assume that verbal coordinations are of different types and that it is possible
to reduce all relevant verbal coordinations to propositional domains. We can conclude
that the G-operator can always be adjoined below the first verbal conjunct without leading
to Presupposition failure.
Notice that we have shifted our reasoning that the difference between nominal and verbal
coordination is related to existence or non-existence of a pronominal counterpart. The
existence of a pronominal counterpart is relevant only in an indirect way. The reason why a
verbal coordination asymmetrically c-commanded by new material is felicitous is not that
there is no better tool to mark givenness. The reason is that such a coordination is a propositional
domain. Thus, the G-operator can be safely inserted. The situation with coordinated
DPs is rather different. There is no way to insert the G-operator without Presupposition
failure. Thus, invoking givenness lexically is the only grammatically suitable option the
language has.
What about adverbs? Recall that adverbs are sensitive to ordering within a coordination
but they do not need to be pronominalized even if there is a pronominal counterpart available.
I argue that adverbs can always be adjoined at a propositional level. Thus, an adverbial
coordination is always located on the left edge of a propositional domain. Therefore,
inserting a G-operator below the first conjunct is always felicitous. There is one problem
remaining: we have defined the reference set as a set containing derivations based on the
same numeration and the same assertion. Presumably, DPs and their pronominal counterparts
are not part of the same numeration. Even though there might be a more principal
solution of the problem, for now I will stipulate that pronouns may be part of the reference
set as well. 18
18 One option would be to treat pronouns as DP ellipsis, as has been suggested in Postal (1966) and further
developed in Elbourne (2005).
119

To sum up, we have seen that the G-operator that we need anyhow in order to mark
presupposition does a lot of useful work for us. Now we have a system that can capture
all the Czech data we have encountered so far. The only tools we need are G-movement,
the G-operator and the assumption that the semantic component is able to compute global
comparison of derivations.
4.5 The interpretation of Givenness
In this chapter I have introduced a recursive G-operator whose purpose is to mark a part
of a structure as presupposed. I have not, however, defined yet what it means to be given
in a technical sense. The leading intuition is that given is something that is salient in the
discourse and which gives rise to a presupposition that must be satisfied for the utterance
to be felicitous in the relevant context.
In principle we could adopt various definitions of givenness. I will adopt here Sauerland
(2005)’s definition of givenness. The relevant lexical entries for type e and type et follow.
(99) Lexical entry for Given of type e:
Given g e ∃x & ∃i.g(i) = x
= λx
x
(100) Lexical entry for Given of type et:
where f has a salient antecedent in C
Given = λf et ∃x ∈ De .f(x) = 1
f
If Given applies to an element of type < α,t >, Given presupposes the existential closure
of the complement of Given. If Given applies to an element of type e, the lexical entry
requires the element to be evaluated with respect to an assignment previously established
in the discourse. As we will see in section 4.6 these lexical entries are too weak but I will
stay with them for lack of a better alternative. To see how these lexical entries combine
with our current system, consider the following example, after Sauerland (2005).
(101) English version:
a. Q: Who ate a cookie?
b. A: LINA [ate a cookie]-Given
(102) Czech version:
a. Q: Kdo snědl koláček?
who ate cookie?
‘Who ate a/the cookie?’
b. A: Koláček snědla || Lina.
cookie ate Lina
‘Lina ate a/the cookie.’
120

In this case, the given part is ‘ate a/the cookie’. Let’s go step by step through the derivation
and its interpretation. A simplified LF of the Czech sentence in (102-b) is given in (103).
Basic lexical entries are given in (104).
(103) E
cookie
D
ate
C
G
B
8 A
7 vP
Lina
VP
(104) a. ate = λx.λy. y ate x
b. Lina = lina
c. cookie = cookie
where lina and cookie are individuals
t7 ate
t8 cookie
Let’s compute the semantics of the LF in (103) step by step, starting with the denotation of
VP and vP.
(105) VP = λy. f 7 (x 8 )(y)
(106) vP g = 1 iff
[λy. f 7 (x 8 )(y)] (Lina) = 1 iff
[λy. f 7 (x 8 )(y)] (lina) = 1 iff
f 7 (x 8 )(lina)
After taking into account lambda abstraction induced by G-movement of the object and the
verb we get the following denotation.
(107) B = λf 7 .λx 8 . f 7 (x 8 )(lina)
Now the G-operator can take the constituent B as its argument because B is not of type t.
The resulting denotation is given in (108). Notice that the G-operator applies in two steps.
First, it induces that there is going to be a given function of type e,et. In the following step,
the operator induces a given individual.
(108) C = G(B) =
G(λf 7 .λx 8 . f 7 (x 8 )(lina)) =
λh e,et : Given(h). G ([λf 7 .λx 8 . f 7 (x 8 )(lina)](h)) =
λh e,et : Given(h).G ([λx 8 . h(x 8 )(lina)]) =
121

λg e .λh e,et : Given (g) & Given(h). G ([λx 8 . h(x 8 )(lina)](g)) =
λg e .λh e,et : Given (g) & Given(h). G (h(g)(lina)) =
λg e .λh e,et : Given (g) & Given(h). h(g)(lina)
This constituent first combines with the verb ate, marking ate as given.
(109) D = C (ate) =
[λg e .λh e,et : Given (g) & Given(h). h(g)(lina)] (ate) =
[λg e .λh e,et : Given (g) & Given(h). h(g)(lina)] (λw.λz. z ate w) =
λg e : Given (g) & Given(ate). lina ate g
In the next step, the structure combines with the object cookie, marking cookie as given.
(110) E = D (cookie) = D(cookie) =
[λg e : Given (g) & Given(ate). lina ate g](cookie) =
Given(cookie) Given(ate): when defined = 1 iff lina ate cookie
Let’s now explicate what Given(c) and Given(ate) mean, using the lexical entry from (99)
and a slightly modified entry from (100). The reason we need to modify the entry is that
the original entry is of type et but ate is of type e,et.
(111) Given g (cookie) =
[λh e : ∃h & ∃i.g(i) = h . h](cookie) =
∃cookie & ∃i.g(i) = cookie . cookie
(112) Given(ate) =
[λf e,et : ∃x ∈ D e .∃y ∈ D e . f(y)(x) = 1 & f has a salient antecedent . f(y)(x)=1]
(ate) =
[λf e,et .∃x ∈ D e .∃y ∈ D e : f(y)(x) = 1 & f has a salient antecedent . f(y)(x)=1 :
f] (λz.λw. w ate z) =
∃w ∈ D e .∃z ∈ D e : w ate z = 1 & f has a salient antecedent . λz.λw. w ate z
We can now insert these two denotations into the original derivation. The reader may wonder
whether we get the right meaning if we treat givenness of ate and cookie separately.
The intuition is that it is the whole constituent ate cookie that is given. The final meaning
is that the sentence in (102-b) is true only if Lina ate a cookie and if the following presupposition
is satisfied: there is someone who ate a cookie and the cookie has been previously
introduced in the discourse. While the denotations combine by functional application, the
presuppositions simply conjoin.
(113) Given(cookie) & Given(ate) when defined = 1 iff lina ate cookie =
(∃cookie & ∃i.g(i) = cookie)&(∃w ∈ D e .∃z ∈ D e : w ate z = 1 & f has a salient
antecedent) when defined = 1 iff lina ate cookie
I leave open for now whether this is the right meaning or whether the presupposition needs
to be strengthened. We know that the weak meaning (conjunction of presuppositions) is
sometimes needed but the open question is whether it is always sufficient.
122

To sum up, we now have in place a formal system which can account both for syntax
and semantics of givenness in Czech. In the following sections, I will address two remaining
questions: First, what is the relation between givenness and deaccenting in English?
Second, what is the role of phonology in the overall system?
4.6 Givenness versus deaccentuation
In the literature on givenness there is disagreement on how to define the exact semantic
properties of givenness. 19 I have not addressed so far the question of how G-movement
stands with respect to the conditions on givenness assumed for other languages. Since the
literature on givenness is vast I will concentrate here on literature which explicitly argues
for a direct relation between deaccentuation in English and givenness. This is a reasonable
move because the literature arguing for givenness as the default information structure value
(Schwarzschild, 1999; Krifka, 2001; Sauerland, 2005, among others) is based on the claim
that an empirically more accurate theory of English prosody can be built if focus accenting
is eliminated in favor of deaccenting given constituents. 20
I will show in this section that the conditions on deaccenting are weaker than the conditions
on G-movement. In other words, not everything that gets deaccented in English,
G-moves in Czech. I will argue that for a Czech constituent α to be given and be able to
G-move two conditions must be satisfied: (i) α must have a salient (contextually entailed)
antecedent, and (ii) there must be an existential presupposition which α satisfies. On the
other hand, in English, for an element to be deaccented it is enough that the element has
a salient antecedent. The main argument will be based on the behavior of indefinites in
Czech.
Let’s start with Schwarzschild’s informal definition of givenness, in (114).
(114) Definition of GIVEN (Schwarzschild 1999, p. 151, (25))
An utterance U counts as GIVEN iff it has a salient antecedent A and
a. if U is type e, then A and U corefer;
b. otherwise: modulo ∃–type shifting, A entails the Existential F-closure of U.
The question that arises is what it exactly means to have a salient antecedent. Consider the
following English examples and the parallel Czech examples.
(115) (Schwarzschild, 1999, p. 154, ex. (30))
If a MAN owns a DONKEY, his WIFE owns a donkey.
(116) a. Když muž vlastní OSLA, jeho žena vlastní OSLA.
when man owns donkey his wife owns donkey
‘If a man owns a donkey, his wife owns a donkey.’
19 I thank Irene Heim for discussing the facts in this section with me and for helping me sort them out.
20 But see Krifka (2006a) and Krifka (2006b) for arguments that not everything can be captured with a
givenness-only theory.
123

. #Když muž vlastní osla, osla vlastní jeho žena.
when man owns donkey donkey owns his wife
(117) In general, CHILDREN misbehave. But teachers only SOMETIMES punish
children.
(118) a. Jezevčíci jsou obvykle nevychovaní, ale jenom někdy
dachshunds.Nom are usually misbehaved but only sometimes
důchodci trestají JEZEVČÍKY.
pensioners.Nom punish dachshunds.Acc
‘Dachshunds usually misbehave but pensioners only sometimes punish dachshunds.’
b. #Jezevčíci jsou obvykle nevychovaní, ale jezevčíky jenom
dachshunds.Nom are usually misbehaved but only sometimes
někdy trestají důchodci.
dachshunds.Acc punish pensioners.Nom
The examples in (115)–(118) contain existential indefinites. As we can see, in English, repetition
of an existential indefinite is a sufficient condition for the indefinite to be deaccented.
In fact, the indefinites must be deaccented. In contrast, in Czech, such an indefinite does
not undergo G-movement. Furthermore, the indefinite carries the main sentential stress.
Deaccentating is not possible.
As the example in (119) demonstrates, in a non-generic context a plural DP like dachshunds
may undergo G-movement. But if it undergoes G-movement then it is necessarily interpreted
as a definite description. If there is no G-movement, as in (120), the second occurrence
of dachshunds must be interpreted as a different group of dachshunds than those who
were misbehaving.
(119) Jezevčíci
dachshunds.Nom
trestají
punish
jsou
are
důchodci.
pensioners.Nom
někdy nevychovaní, a pak jezevčíky
sometimes misbehaved and then dachshunds.Acc
a. ‘Dachshunds sometimes misbehave and then pensioners punish the dachshunds.’
b. #‘Dachshunds sometimes misbehave and then pensioners punish dachshunds.’
(120) Jezevčíci jsou někdy nevychovaní, a pak důchodci
dachshunds.Nom are sometimes misbehaved and then pensioners.Nom
trestají jezevčíky.
punish dachshunds.Acc
a. ‘Dachshunds sometimes misbehave and then pensioners punish dachshunds.’
b. #‘Dachshunds sometimes misbehave and then pensioners punish the dachshunds.’
To summarize so far, existential indefinites do not G-move in Czech and they have no problem
being accented in a position where an English existential indefinite must be deaccented.
On the other hand, if a plural (recall that Czech does not have overt articles) undergoes G-
movement, it must be interpreted as a definite description.
124

Generic indefinites behave slightly differently than existential indefinites. They do not
undergo G-movement either, but they may be deaccented if something else needs to bear
the main sentential stress. Consider the following examples containing hyponyms.
(121) (Ladd, 1980)
Q: Has John read Slaughterhouse-Five?
A: He doesn’t READ books.
(122) a. On knihy NEČTE.
he books not-reads
b. On NEČTE knihy.
he not-reads books
‘He doesn’t READ books.’
(123) (Chafe, 1976)
I bought a painting last week.
I really LIKE paintings.
(124) a. Já mám obrazy RÁD.
I have paintings liked
b. Já mám RÁD obrazy.
I have liked paintings
‘I really LIKE paintings.’
In Czech as in English, the hyperonym primed by the previous context is deaccented. If we
look more closely at the examples, we see, however, that they are rather different. Let’s first
consider the example in (121). The answer in (121) is a correction to a proposition entailed
by the question. The question about reading Slaughterhouse-Five entails an answer about
reading books. 21 Corrections in general assert that an entailed (or asserted) proposition
is not true. As a result, in this particular example, the negation carries focus accent as a
means of excluding the entailed proposition from the set of propositions compatible with
the common ground.
On the other hand, the example in (123) does not contain any correction. The relevant
utterance is an answer to an implicit question ‘Why do you buy paintings?’. 22 The implicit
question primes the focus accent on the evaluative predicate like. Even though these two
cases are not identical, they both contain a focused element that needs to be stressed. I
argue that the fact that the generic indefinites are deaccented is not a property of the indefinites
per se but is a consequence of their prosodic neighbor being stressed.
If this conclusion is correct, we predict that if there is no focus, hyponyms cannot be
deaccented. As we can see in (125), this is indeed correct. Thus, the fact that the generic
indefinites in the previous examples are deaccented has nothing to do with them having a
salient antecedent. Instead, the lack of sentential stress is a side-product of focusing the
21 This follows only if an answer has the same implicature as the question.
22 I thank Danny Fox, p.c., for suggesting that I treat these cases as containing an implicit question.
125

immediately preceding prosodic word.
(125) My son likes gorillas.
a. In general, he likes ANIMALS.
b. #In general, he LIKES animals.
In section 4.7, I will argue that the short movement we observe in Czech is also the result
of a prosodic requirement and has nothing to do with givenness. For now let’s consider the
following example which shows that whatever triggers the displacement of the object in
(122-a) and (124-a) is not G-movement. As we have seen in the previous chapters, a given
element cannot be asymmetrically c-commanded by a new element within a propositional
domain. If this is the case, the given element must move. As we can see in the example
in (126), in case of generic indefinites accompanied by a focused element, the indefinites
move only locally. Whether there is any new element or not is irrelevant for the movement.
Thus, we can conclude that this kind of movement is not G-movement.
(126) a. I bought a painting last week.
b. Moje nová přítelkyně obrazy SBÍRÁ.
my new girlfriend paintings collects
‘My new girlfriend COLLECTS paintings.’
The question is whether indefinites ever G-move. We have already seen several examples
suggesting the opposite, but there seem to be some counterexamples. See, for instance, the
example in (127) and (128).
(127) Q: I’ve heard you finally sold all the used cars from your store. You also had
several Porsches, right? Do you remember who bought a Porsche?
A: Yes. For example, a FRIEND OF MY WIFE bought a Porsche.
(128) a. Porsche si například koupil kamarád mojí
Porsche REFL for example bought friend of-my
‘For example, a friend of my wife bought a Porsche.’
b. #Kamarád mojí ženy si například koupil
friend of-my wife REFL for example bought
ženy.
wife
Porsche.
Porsche
The example in (128) contrasts with the example in (130). In both utterances there is a
salient antecedent but only in (128) does the indefinite move. What is the relevant difference?
I want to suggest that the indefinite in (128) is semantically a partitive (a Porsche
from the Porsches in the store). As such it triggers an existential presupposition, in contrast
to the indefinite in (130) which cannot be interpreted as a partitive and therefore cannot
give rise to an existential presupposition.
(129) Q: Do you happen to know someone who owns a Porsche?
A: Yes, a FRIEND OF MY WIFE owns a Porsche.
(130) a. Kamarád mojí ženy má Porsche.
friend of-my wife has Porsche
‘A friend of my wife has a Porsche.’
126

. #Porsche má kamarád mojí ženy.
Porsche has friend of-my wife
Notice that the partitive interpretation brings about a definite interpretation. There is nothing
in our semantics of givenness that would give us the partitive interpretation. The partitive
semantics seems to be independent of givenness even though it is a precondition for
an indefinite to be treated as given. I leave for future research how exactly givenness and
partitivity interacts.
Another problem that I leave unanswered is how we can restrict the existential presupposition
only to the relevant quantificational domain. Presumably, it is part of our common
background that Porsches exist but it is not enough for a Porsche to be given.
An additional example showing the relation between an existential presupposition and
the inability of an indefinite to undergo G-movement is given in (132).
(131) I have entered Irene’s office and I see a Porsche key on her desk. I ask:
a. Q: Who bought a Porsche?
b. A: KAI bought a Porsche.
(132) a. Kai si koupil
Kai REFL bought
‘Kai bought a Porsche.’
b. #Porsche si koupil
Porsche REFL koupil
‘Kai bought a Porsche.’
Porsche.
Porsche
Kai.
Kai
Why is there no movement in this example? Notice that there is a crucial difference between
the example in (128) and the example in (132). In example (132) the existence of a
Porsche is only entailed from the context. It has not been asserted that there is a Porsche
that someone bought. Irene could have felicitously answered, for instance, ‘There is no
Porsche. I found the key.’
Based on the examples with indefinites, I argue that having a salient antecedent is not
a sufficient condition for treating an element in Czech as given. The condition must be
strengthen by a requirement that a given element must give rise to an existential presupposition
which is already part of the common ground. The new definition is given in (133).
(133) Definition of GIVEN for Czech
An utterance U counts as GIVEN iff
a. it is presupposed that U exists, and
b. it has a salient antecedent A
(i) if U is type e, then A and U corefer;
(ii) otherwise: modulo ∃–type shifting, A entails the Existential F-closure
of U.
Adding this condition explains why so many of the examples we have seen so far were
127

translated by using a definite description. At the same time keeping the original requirement
on a salient antecedent ensures that not every definite description will be treated as
given. Consider the example in (134).
(134) Král včera hrál šachy s královnou.
king yesterday played chess with queen.
‘The king and the queen played chess yesterday.’
The example in (134) is a felicitous answer to the question ‘What happened?’ in a context
where there is only one king and one queen in a palace. Even though ‘the queen’ is interpreted
as definite, there is no G-movement. For ‘the queen’ to be able to G-move, there
would have to be a salient antecedent like in the example in (135).
(135) a. The queen was bored all day but then. . .
b. s královnou hrál král šachy.
with queen played king chess
‘. . . the king played chess with the queen.’
To summarize this section, I have provided an argument that, in Czech, for an element to be
given (at least for purposes of G-movement) it is not enough to have a salient antecedent.
A stronger semantic condition is required. Specifically, I have argued that for an element
to be interpreted as given the element must give rise to an existential presupposition. Thus,
the semantic condition on being given in Czech is stronger than the condition on being
deaccented in English.
4.7 A note on phonology and its relation to givenness
Superficially, Czech behaves like languages that require a new element to occupy the most
embedded part of the tree in order to be assigned nuclear stress (Cinque, 1993; Zubizarreta,
1998; Arregui-Urbina, 2002; Szendrői, 2003). One may wonder whether we are losing
something if we do not refer to phonology at all in the treatment of givenness in Czech.
In this section, I will briefly address this question. I will first review the basics of Czech
sentential prosody. Then, I will show a couple of cases where elements move in order to
satisfy their prosodic requirement. I will argue that this kind of movement is different from
G-movement. Finally, I will briefly address the question of abstract phonology as a trigger
of syntactic movement.
Main sentential prominence in Czech is realized on the first syllable of the last prosodic
word (cf. Gebauer 1900; Skaličková 1956; Daneš 1957; Romportl 1973; Palková 1994,
among others). 23 Czech, in contrast to English, does not realize sentential prominence with
a high tone. Instead, prominence is realized by increased intensity and longer duration. 24
23 For further discussion on the Czech sentential prosody and its relation to information structure see Mathesius
1931, 1937; Trávníček 1937, 1939; Petřík 1938; Petr et al. 1986, among others.
24 It is, however, possible that what is perceived as prosodic prominence is a high tone on the stressed
syllable of the last prosodic word followed by a low tone (Edward Flemming, p.c.).
128

The position of the sentential stress falls on the main word stress. Main word stress falls
on the initial syllable of a prosodic word. In general, odd-numbered syllables are stressed,
giving rise to trochaic rhythm. For an overview of word stress in Czech see van der Hulst
1999, p. 818ff.
Ladd 1996 argues that Czech is a language that does not use pitch accent at all and that
has very rigid phonology. This is significant in light of the hypothesis that syntax must
adjust to phonology because phonology is not a ‘plastic’ component of the grammar (cf.
Vallduví 1990). 25 This picture is, however, too simplistic. Czech has other prosodic means
than assigning nuclear stress to the last prosodic word. For example, contrastively focused
words may be prosodically prominent in situ. Also certain lexical items, for instance whwords,
are independently prosodicaly prominent. But it is true that Czech prosody is more
rigid than the prosody of, for example, English. Thus, we cannot exclude in advance the
possibility that syntax accommodates prosodic requirements.
The important question with respect to givenness is whether there is a direct relation
between being given and deaccenting. We have already seen in the previous section that
the semantic licensing conditions on givenness are stronger than the licensing conditions
on deaccenting in English, but there is still a possibility that for an element to be given in
Czech, the element must be deaccented. Thus, the question is whether or not G-movement
may be characterized as prosodically driven movement.
To be able to answer this question, let’s first identify another type of prosody driven
movement in Czech. Such movement is found with inherently stress-less items, for instance
něco ‘something’. As we can see in (136), ‘something’ cannot be the last prosodic
word. The reason is that in this position it would bear the main sentential stress. Crucially,
destressing ‘something’ in situ is not possible.
(136) a. *Chtěla bych jíst něco.
wanted would eat.Inf something
b. Chtěla bych něco jíst .
wanted would something eat
‘I would like to eat something.’
As we can see in the example in (137), the movement observed in (136) is not independent
syntactic movement. If we modify the sentence in (136) with a VP adverbial which
surfaces as the last prosodic word, ‘something’ does not need to move anymore. In fact, it
cannot move. The reason is that now it is the adverbial that bears the main sentential stress.
‘Something’ is not affected by the sentential prominence.
(137) a. Chtěla bych jíst něco u Nováků.
wanted would eat.Inf something at Novak’s
b. *Chtěla bych něco jíst u Nováků.
wanted would something eat at Novak’s
25 Czech is supposed to pattern in this respect with languages like Italian and Hungarian.
129

‘I would like to eat something at Novak’s.’
Furthermore, other lexical items that are semantically similar but at the same time can be
stressed, such as ‘some food’, do not move, as can be seen in (138).
(138) a. Chtěl bych jíst nějaké jídlo.
wanted would.1sg eat.Inf some food
‘I would like to eat some food.’
b. Chtěl bych jíst polévku.
wanted would.1sg eat.Inf soup
‘I would like to eat some soup.’
To summarize the observations so far, in Czech sentential prominence is realized only on
the last prosodic word. For an element to be deaccented it is enough to move very locally.
No cyclic movement is required (or licensed).
I have already mentioned that in certain cases it is possible to realize prominence in a
different part of the structure as well. For instance, if something is contrastively stressed,
the immediately preceding prosodic word is deaccented. Thus, we predict that a stress-less
word like ‘something’ does not need to evacuate the last position if the preceding word is
contrastively stressed. This is indeed correct, as we can see in (139).
(139) Chci JÍST něco. Ne, PÍT něco.
want-I to-eat something no to-drink something
‘I want to EAT something, not to DRINK something.’
In general, there is an optionality in whether the contrastive element is stressed in situ or
whether the contrastive element is realized as the last prosodic word, as in (140).
(140) Chci něco JÍST, ne PÍT.
want-I something to-eat no to-drink
‘I want to EAT something not to DRINK something.’
The examples in (139) and (140) are similar to the deaccented examples with hyponyms
from the previous section. Consider again the example from (124) repeated below as (141).
If the contrasted predicate ‘like’ stays in situ, as in (141-b), it gets stressed and the following
prosodic word ‘paintings’ get deaccented. On the other hand, if the object ‘paintings’
undergoes short movement, the predicate is assigned the main sentential stress.
(141) a. Já mám obrazy RÁD.
I have paintings liked
b. Já mám RÁD obrazy.
I have liked paintings
‘I really LIKE paintings.’
Notice that Czech does not seem to have the option of deaccenting an element by shifting
the main sentential stress. The only way something can become deaccented is if its immediately
preceding prosodic neighbor is independently stressed. The question that arises
130

is whether English givenness is really so different from Czech. As far as I can tell, it is
quite possible that deaccenting by shifting stress is licensed under the same conditions as
G-movement in Czech. Other deaccenting cases might be a result of destressing caused by
stressing of the immediately preceding prosodic neighbor. But whether this is a plausible
theory or not must be left for future research.
To conclude, we have identified a type of movement which is purely driven by prosodic
requirements. We can now ask the question whether G-movement is the same type of
movement or not. If G-movement is motivated by prosody in the same way as the movement
observed above, then there are two logical options to be considered: either (i) given
elements cannot be stressed, i.e., they move in order to not bear nuclear stress, or (ii) new
elements move to the position in which they would be marked by nuclear stress.
Neither of these two options seems to be correct. We know from the behavior of ‘something’
that for a word not to be stressed it is enough to move very locally. G-movement, on
the other hand, may be cyclic and non-local, thus, option (i) cannot be correct. Similarly, if
a new element needs to bear a nuclear stress, there is only one position that is good enough.
Again, we would expect different behavior than the behavior we find. We would expect
that any structure in which one new element would be the last prosodic word should be
well formed, which is not correct. To conclude, option (ii) cannot be maintained either.
The more interesting question is whether a prosodic account of givenness in Czech
could be maintained if we considered a more sophisticated abstract version of the nuclear
stress rule. My worry is whether it is possible to establish such a rule without referring to
the semantic properties of the relevant elements. Recall from our previous discussion that
the partition between given and new may appear in different parts of the tree. Furthermore,
it is sensitive to the propositional complexity of an utterance. Once we need to refer to
the semantic values then it is not clear what the advantage of having a prosodically driven
system would be. The system would necessarily double the work of the semantic and pragmatic
components.
Recent accounts of information structure in terms of the syntax-phonology interface,
such as Selkirk (1995), Reinhart (1995, 2006) or Szendrői (2003), do indeed use some
form of semantic diacritics that they introduce into the derivation. Even though my account
is very close to the syntax-prosody accounts, it is simpler in that it minimizes the
inter-modular interaction. The only thing we need is syntax generating free movement
and semantics choosing from syntactically available structures. The only task left for
the phonology interface is to read off the syntactic interpretation by using independently
needed prosodic rules. 26
26 Such a prosodic system is for example that of Wagner (2005).
131

4.8 Summary
This chapter completed the previous discussion of the syntax of G-movement by creating a
system of semantic interpretation which both improves the coverage of the syntactic system
and provides an explanation for the empirical generalizations made in the previous chapters.
In particular, I argued that G-movement is movement with semantic consequences in
the sense that syntactic items move in order to create a linear partition between elements
to be interpreted as given and elements to be interpreted as new. We saw that the same
partition can be established not only by G-movement but also by base generation as in the
case of coordinations and basic word orders. I argued that the purpose of the partition is to
create a configuration in which it is possible to insert a given operator.
I argued that in order to account for the Czech data we need a recursive operator which
applies to a non-propositional constituent and turns its sister into a presuposed element.
I proposed that the distribution of the operator should be governed within a comparative
system which compares derivations with respect to the Maximize Presupposition maxim of
Heim (1991). I proposed that the evaluation is done over a set which contains derivations
based on the same numeration and assertion. We saw that the combination of the given
operator and the evaluation component captures all the relevant data.
In the final two sections I addressed the question of how G-movement relates to deaccenting
for givenness in English. I showed that the licensing conditions on G-movement are
stronger in that they require that, for an element to be given, it must not only have a salient
antecedent but also be existentially presupposed. In the last section I shortly commented
on a worry that we might be losing an important generalization if we do not give any role
to phonology in our system. I argued that this move is indeed harmless: with respect to
givenness, the phonology component behaves only as an interpretative component.
132

Appendix A
G-movement is A-movement
To complete the overall picture of the properties of G-movement I will address in this
appendix the question of whether G-movement is A-movement or A-bar movement. In
particular, I will argue that G-movement is A-movement in the sense that it creates a new
binding configuration. But first I will define what I mean by the A- versus A-bar distinction.
There has been disagreement in the literature about classifying scrambling as A-movement
or A-bar movement. The disagreement lies in how we define the distinction between A-
and A-bar movement in the first place. For approaches that base the distinction on Case assignment,
there are two basic options: Either (i) only a DP that is assigned Nominative can
move to Spec, TP since T assigns Nominative Case; thus, scrambling of a non-Nominative
DP is by definition movement to a position higher than Spec,TP and must be an instance
of A-bar movement (cf. for example Baker (2003)). Or, (ii) a position like Spec,TP may
be ambiguous between A-movement and A-bar movement depending on whether the DP
that moves there is or is not nominative (see, for example, Bonet (1990); Diesing (1990);
Mahajan (1990)).
On the other hand, it has been argued that A movement and A-bar movement behave
differently with respect to binding (Lebeaux (1988, 1998); Saito (1989); Mahajan (1990);
Chomsky (1993, 1995); Fox (1999); Lasnik (1999)). I will adopt here the binding distinction
over the Case distinction. I will classify as A-movement only movement, where the
pronounced copy is relevant for binding, and as A-bar movement movement where the pronounced
copy is not relevant for binding. 1 It is outside of the scope of this work to explain
why there should be such a difference between A- and A-bar movement. 2
In particular, I will follow recent proposals for Russian (see Lavine and Freidin (2002);
Bailyn (2003, 2004), among others) that movement to Spec,TP in Russian, even if it is
movement of an object, is always A-movement, i.e., it creates a new binding configuration.
1 I simplify the discussion here. It has been argued that A-movement can reconstruct as well (see, for example,
an overview in Iatridou 2002). The question of reconstruction is immaterial for the current discussion.
The only relevant difference is whether the pronounced copy counts for binding or not.
2 One option is that being A- or A-bar follows from a feature composition of a particular projection, not
from the type of a projection per se. See Nevins and Anand 2003 for an idea in this direction.
133

As we will see below, the same holds for G-movement. The examples in the following
sections will partially be modeled after Lavine and Freidin (2002); Bailyn (2004).
Notice that once we tie the A- versus A-bar distinction only to binding, it becomes
non-trivial to distinguish between movement and base generation. If G-elements were base
generated on the left periphery, i.e., they would be something like clitic left dislocation
in Romance (for example, Cinque 1990; Iatridou 1995; Barbosa 1995), we would not be
able to tell from the binding interactions. I will comment on base generation in section A.4.
The argument for A-movement properties of G-movement to Spec,TP is based on binding
interactions, in particular on Condition A, Condition C, and Weak Cross-Over effects.
I am not using other tests for absence of A-bar movement used in Cinque 1990; Iatridou
1995; Barbosa 1995. The reason is that most of the tests are based on the absence or the
presence of a clitic. There is no clitic in G-movement constructions in Czech. However,
as we will see in section A.4 there are constructions that are more similar to clitic left dislocation
in Romance. For closer discussion of these and their relation to contrastive left
dislocation see Sturgeon 2006. Thus, in the following sections I will argue for (1).
(1) G-movement is A-movement.
A.1 Condition A
Let’s start by looking at Condition A. The condition is given in (2).
(2) Condition A:
A reflexive (or reciprocal) pronoun must be bound within its local domain.
If G-movement is A-movement, G-movement of a DP containing a reflexive pronoun over
its binder, should lead to a Condition A violation. On the other hand, we predict that if
G-movement is A-bar movement, it should not affect the ability of the reflexive pronoun to
be bound.
Even though the predictions are clear, testing for Condition A in Czech is not a straightforward
task. The reason is that Czech has anaphoric reflexive pronouns that are always
bound by a Nominative argument. As the following examples show, binding of reflexive
pronouns is independent of the actual structural relation between the anaphor and its
binder. 3
(3) Nom-Acc verbs:
a. Marie má ráda svoji kočku.
Marie.Nom-i has liked her-i cat
3 Examples with sebe ‘oneself’ are not presented as minimal pairs. The reason is that the given counterpart
of sebe is lexicalized as a reflexive second position clitic that never occupies Spec,TP. Focusing the object by
association with jenom ‘only’ gives us the desired interpretation in a way that avoids the problem with using
the clitic.
134

. Svoji kočku má ráda Marie.
her-i cat.Acc has liked Marie.Nom
‘Marie likes her cat.’
(4) Nom-Locative:
a. Petr bydlel ve svém domě.
Petr.Nom-i lived in his-i house.Loc
b. Ve svém domě bydlel jenom Petr.
in his-i house lived only Petr-i
‘(Only) Petr lived in his own house.’
(5) Nom-Acc:
a. Petr viděl sebe.
Petr.Nom saw himself
b. Sebe viděl jenom Petr.
himself saw only Petr
‘(Only) Petr saw himself.’
(6) Ditransitive verbs:
a. Petr dal svému bratrovi knihu.
Petr.Nom-i gave his-i brother.Dat book.Acc
b. Svému bratrovi dal Petr knihu.
his-i brother.Dat gave Petr.Nom book.Acc
‘Petr gave his brother a book.’
(7) Ditransitive verbs:
a. Petr dal bratrovi svou knihu.
Petr.Nom-i gave brother.Dat his-i book.Acc
b. Svou knihu dal Petr bratrovi.
his-i book.Acc gave Petr.Nom brother.Dat
‘Petr gave his book to (his) brother.’
Fortunately, the facts are different if the binder is a quirky subject, i.e., a Dative argument.
In such a case, binding is possible only if the Dative is in the subject position (Spec, TP);
otherwise, violation of Condition A arises. 4
(8) Dative subject:
a. ?Máše bylo líto svého psa.
Máša.Dat-i was sorry her-i dog.Gen
b. *Svého psa bylo líto Máše.
her-i dog.Gen was sorry Máša.Dat-i
‘Máša felt sorry for her dog.’
(9) Dative subject:
4 To my knowledge, the asymmetry between Nominative and Dative subject has not been reported in
literature yet. I have no explanation for the binding difference.
135

a. Máši bylo líto sebe.
Máša.Dat was sorry herself.Gen
b. *Sebe bylo líto Máši.
herfself.Gen was sorry Máša.Dat
‘Máša felt sorry for herself.’
We see that G-movement causes a Condition A violation, suggesting that G-movement is
A-movement.
Further evidence comes from possessive pronouns that are anaphorically bound by a
non-subject element but are not morphologically marked as reflexive. Consider the examples
in (10) and (11). 5
(10) a. *Její kočka má ráda Marii.
her-i cat.Nom has liked Marie.Acc-i
b. Marii má ráda její kočka.
Marie.Acc-i has liked her-i cat.Nom
‘Mary is loved by her cat.’
(11) a. ??Jeho přátelé obdivují Petra.
his-i frinds.Nom admire Petr.Acc-i
b. Petra obdivují jeho přátelé.
Petr.Acc-i admire his-i friends.Nom
‘His friends admire Petr.’
As we can see, such a possessive pronoun is felicitous only if the R-expression coindexed
with the pronoun undergoes G-movement over the pronoun. Thus, the R-expression is in
a position from which it can bind. It follows that the position must be an A-position and
G-movement is A-movement. 6
A.2 Condition C
This section will comment on Condition C. The Condition C is defined in (12).
(12) Condition C:
An R-expression (= proper name or definite description) must be free.
What kind of interactions do we expect with respect to G-movement and Condition C?
First of all we need to establish that Czech is sensitive to Condition C. I will show this
5 There is a difference in the level of unacceptability between the examples in (10) and (11). David
Pesetsky, p.c., suggested that the difference in acceptability might come from inaliable versus non-inaliable
possession relations. At this point I do not have an explanation for the gradient differences in judgments.
For now, however, the crucial point is that there is a consistent difference depending on c-command relations
between relevant pairs of arguments.
6 As David Pesetsky, p.c., pointed out, the argument is not conclusive because the same result in grammaticality
would be captured by a theory relying only on linear precedence. Even though this objection is
relevant with respect to Condition A, it does not apply to other binding facts presented at this section.
136

with movement of a pronoun over a coindexed R-expression. If such a pronoun moves
over a coindexed R-expression, it should cause a Condition C violation. The reason is that
A-movement creates a new binding position. If the pronoun undergoes A-movement, it
binds from its surface position. On the other hand, if the pronoun undergoes A-bar movement,
Condition C would not be violated because the pronoun would be able to reconstruct.
I will demonstrate the interaction with OVS examples. Before we approach the actual
test, a note on an OVS order is required. In an OVS order, there are two available parses for
the position of the object. (i) the object can occupy an A-position, or (ii) it can occupy an
A-bar position. I argue that these two positions differ in two aspects: (i) they have different
intonational contours (an element in an A-bar position is contrastively stressed and the
following material gets deaccented; no such change in the intonational contour is observed
with A-position); (ii) in cases without contrastive stress and deaccenting the verb must
linearly precede the subject; as for the other type, the verb can either precede or follow the
subject (capitalized letters mark contrastive stress). Thus, if there is an optionality between
OVS and OSV order then the object could have not undergone A-movement.
(13) Non-contrastive intonation:
a. Petra viděla Marie.
Petr.Acc saw Marie.Nom
b. *Petra Marie viděla.
Petr.Acc Marie.Nom saw
‘Marie saw Peter.’
(14) Contrastive intonation:
a. PETra viděla Marie.
Petr.Acc saw Marie.Nom
b. PETra Marie viděla.
Petr.Acc Marie.Nom saw
‘It was Peter whom Marie saw.’
I suggest that the non-contrastive type corresponds to A-position, i.e., Spec,TP. The obligatory
OVS order is a result of moving the object to Spec, TP and leaving the subject lower in
the structure (presumably within vP; for a similar argument see, for example, Bonet 1990;
Diesing 1990; Miyagawa 1997; Bailyn 2003, 2004). On the other hand, the contrastive
pattern corresponds to A-bar movement, i.e., to movement to a position higher than Spec,
TP. 7 Spec, TP is then still available for the subject to move into, resulting into OSV order
that is not available otherwise. Notice that the subject movement is optional because the
EPP-like requirement may be satisfied by the A-bar moved argument but the subject might
still have an independent reason to move (for example, for scope).
7 I have not been able to pinpoint the exact position of the contrastive element: it is plausible that a
contrastive element is adjoined to TP or that there is a syntactic position between TP and CP that can be used
as a landing site for contrastive elements.
137

With this contrast in place, we can test Condition C. The following examples correspond
to the left-most element being in an A-position, i.e., the intonation is not contrastive
and the OSV order is excluded. All degraded sentences get improved if the linearly first
element gets contrastively stressed, i.e., if it occupies, as I argue, an A-bar movement position:
(15) a. Noví Petrovi známí ho představili řediteli.
new Petr’s-i friends.Nom him.Acc-i introduced director.Dat
b. *Jeho představili noví Petrovi známí řediteli.
him.Acc-i introduced new Petr’s-i friends.Nom director.Dat
‘Petr’s new friends introduced him to the director.’
The relevance of the example in (15) is supported by examples with long distance movement.
Long distance movement in Czech is always contrastively stressed, i.e., it always targets
an A-bar position. Unsurprisingly, in such a configuration, Condition C is not violated
because the moved element may reconstruct.
(16) JEHO, chceme, aby noví Petrovi známí představili řediteli.
him.Acc.TOP-i want-we that new Petr’s-i freinds.Nom introduced director.Dat
‘We want Petr’s new friends to introduce HIM to the director.’
To make the argument sound I need to show that reconstruction from an A-bar position
is in fact obligatory. An optional reconstruction would be compatible with the view that
A-movement can, under certain circumstances, reconstruct as well. As we can see in the
next example, this is correct. If ‘Petr’s book’ had the option to not reconstruct, this option
could be a strategy to avoid Condition C violation. Such a strategy is not available, though.
(17) *PETROvu knihu ho Marie prosila spálit.
Petr’s-i book,Acc him.Acc-i Marie.Nom to-burn
‘Marie begged Peter to burn his book.’
We have seen that Czech is sensitive to a Condition C violation and we can now approach
testing G-movement. Since pronouns do not undergo G-movement, we must look at an
opposite relation, i.e., cases where an R-epxression crosses a coindexed pronoun. If in the
original structure, coindexation of the pronoun and the R-expression causes a Condition
C violation, we predict that G-movement of the R-expression might save the Condition C
violation in case it is A-movement. In contrast, if G-movement is A-bar movement we do
not expect to find any improvement with respect to Condition C.
As we can see in the examples in (18), G-movement can indeed save a Condition C
violation. Thus, we can conclude that, with respect to Conditon C, G-movement behaves
as A-movement.
(18) a. *Marie a on i viděli Petrovi i přátele.
Marie.Nom and he saw Petr’s friends.Acc
‘*Marie and he i saw Petr’s i friends.’
b. Petrovi přátele viděla Marie a on i .
138

Petr’s friends.Acc saw Marie.Nom and he
‘*Marie and he i saw Petr’s i friends.’
A.3 The Weak Cross-Over Effect
Finally, let’s look at the Weak Cross-Over effect (WCO), defined in (19).
(19) Weak Cross-Over:
A pronoun can only be bound from an argument position.
As can be seen in (20), no WCO violation arises in case of G-movement of ‘every girl’.
The example in (21) is here as a control. Long-distance movement leads to a WCO violation.
The contrast shown in (20) and (21) supports the argument that G-movement is
A-movement because it creates new binding relations.
(20) a. *Její pes miluje každou holčičku.
her-i dog.Nom loves every girl.Acc-i
b. Každou holčičku miluje její pes.
every girl.Acc-i loves her-i dog
‘Every girl is loved by her dog.’
(21) *KAŽdou holčičku, chci, aby miloval její pes.
every girl.Acc.TOP-i want-I that loved her-i dog
‘I want every girl to be loved by her dog.’
A.4 A note on base generation
So far I have been assuming that the linear partition between given and new elements is
achieved via movement. However, it has been suggested in the literature that elements on
the left periphery that do not have A-bar properties are base generated in their surface position.
This argument has been made in detail for Romance left clitic dislocation structures
(Cinque 1990; Iatridou 1995; Barbosa 1995). Even though these structures show sensitivity
to syntactic islands, i.e., they appear to undergo movement, the above cited authors argue
that the island sensitivity is a result of the presence of a binding chain that is either entirely
independent of movement, or arises from a short movement within the left periphery. One
could argue that Czech G-movement is just another instance of base generation (notice
that the left dislocated elements in clitic left dislocation constructions are given as well)
and that the various word orders are base generated as they are. Another line of such an
argument has been made in connection with scrambling by Haider (1988); Bayer and Kornfilt
(1994); Kiss (1994); Neeleman (1994); Neeleman and Reinhart (1998), among others.
In this section, I present two objections to a base-generation approach. The first objection
is that the base generation hypothesis does not make clear predictions about the final
word order and it would have to be combined with another system that would determine
the final word order. We have seen in chapter 2 and 3 quite a few examples that suggest a
139

derivational nature of the reorderings found in Czech. Furthermore, it is not clear how to
model the relation between head movement and the locality restriction on G-movement in
a base generation system. I take the previous discussion to have made this point sufficiently.
Another point is that Czech has, aside from G-movement, structures that are at least
superficially similar to clitic left dislocation structures in Romance and Greek.
(22) a. Whom did Mary see drunk?
b. Petra (*ho) viděla
Petr.Acc him.Acc saw
‘Marie saw Petr i drunk i .
Marie
Marie.Nom
opilého.
drunk.Acc
(23) Ten Petr, Marie *(ho) viděla opilého.
the Petr.Nom Marie him.Acc saw drunk.Acc
‘As to the Petr i , Marie saw him i drunk i .’
As we can see in (22), in the case of G-movement, the given DP Petra is accusative and
there is no occurrence of a clitic. In contrast, the same argument appears in (23) in Nominative
case (default case). Accusative case is assigned to a clitic. The clitic is obligatory
in this structure. 8,9 Another difference between G-movement and clitic left dislocation in
Czech can be found in embedded contexts. While structures with G-movement can be
freely embedded, as seen in (24), clitic left dislocation is possible only in a matrix environment,
(25).
(24) Maminka říkala, že Petra (*ho) viděla Marie opilého.
mother.Nom said that Petr.Acc him.Acc saw Marie.Nom drunk.Acc
‘My mother said that Marie saw Petr i drunk i .
(25) a. *Maminka říkala, že ten Petr, Marie *(ho) viděla opilého.
mother.Nom said that the Petr.Nom Marie him.Acc saw drunk.Acc
b. *Maminka říkala, ten Petr, že Marie *(ho) viděla opilého.
mother.Nom said the Petr.Nom that Marie him.Acc saw drunk.Acc
c. *Ten Petr, maminka říkala, že Marie *(ho) viděla opilého.
the Petr.Nom mother.Nom said that Marie him.Acc saw drunk.Acc
‘As to the Petr i , my mother said that Marie saw him i drunk i .’
I suggest that the Case differences and the status of a clitic, combined with the restrictions
on the embedding of clitic left dislocation structures, support the hypothesis that
G-movement is indeed movement. As we have seen, Czech has other syntactic strategies
that are more similar to Romance or Greek clitic left dislocation but G-movement is not
one of those.
8 More details on this structure and other left dislocation strategies in Czech can be found in Sturgeon
2006.
9 It has been argued that clitic left dislocation does not always require a clitic (if the left-dislocated element
is the subject or an adjunct) (Cinque, 1990). I control for this possibility by dislocating only objects.
140

A.5 Summary
To conclude, we have seen that G-movement can be characterized as A-movement in the
sense that the position to which element α G-moves functions as a new binding position.
To support the argument, I have provided examples showing interactions between
G-movement and Condition A, Condition C and Weak Cross-Over effect. In the final
section, I have briefly addressed the question whether the reordering I have attributed to
G-movement might be in fact base generated.
141

Bibliography
Aboh, Enoch Oladé. 2004. The morphosyntax of complement-head sequences: clause
structure and word order patterns in Kwa. New York: Oxford University Press.
Adamec, Přemysl. 1962. K úloze sémantiky ve slovosledu [On the role of semantics in
word order]. AUC Philologica, Slavica Pragensia IV:297–300.
Adger, David. 2003. Core Syntax: A Minimalist Approach. Oxford University Press.
Arregui-Urbina, Karlos. 2002. Focus on Basque movements. Doctoral Dissertation, MIT.
Bailyn, John Frederick. 2001. On scrambling: A reply to Bošković and Takahashi. Linguistic
Inquiry 32:635–658.
Bailyn, John Frederick. 2003. Does Russian scrambling exist? In Word order and scrambling,
ed. Simin Kirimi. Malden, MA: Blackwell Publishers.
Bailyn, John Frederick. 2004. Generalized inversion. Natural Language and Linguistic
Theory 22:1–49.
Baker, Mark C. 2003. Agreement, dislocation and partial configurationality. In Formal
approaches to function, ed. A. Carnie, H. Harley, and M. Willie, 107–134. Amsterdam:
John Benjamins.
Bale, Alan C. To appear. Quantifiers and verb phrases: An exploration of propositional
complexity. Natural language and linguistic theory .
Barbosa, Maria do Pilar Pereira. 1995. Null subjects. Doctoral Dissertation, MIT.
Bayer, Josef, and Jaklin Kornfilt. 1994. Against scrambling as an instance of Move-alpha.
In Studies on scrambling, ed. Norbert Corver and Henk van Riemsdjijk, 17–60. Berlin,
New York: Mouton de Gruyter.
Beck, Sigrid, and Uli Sauerland. 2000. Cumulation is needed: A reply to Winter (2000).
Natural language semantics 8:349–371.
Belletti, Adriana, ed. 2004. Structures and beyond. The cartography of syntactic structures,
volume 3. Oxford: Oxford University Press.
Benešová, Eva. 1968. O sémantickém charakteru českého slovosledu [On the semantic
character of Czech word order]. Slovo a slovesnost 29:34–41.
142

Biskup, Petr. In preparation. The phase model and adverbials. Doctoral Dissertation,
Universität Leipzig.
Bonet, Eulàlia. 1990. Subjects in Catalan. In Papers on wh-movement. MIT Working
Papers in Linguistics, ed. Lisa L.S. Cheng and Hamida Demirdash, 1–26. Cambridge,
MA: MITWPL.
Bošković, Željko. 2001. On the Nature of the Syntax-Phonology Interface. Cliticization
and Related Phenomena. Amsterdam: Elsevier.
Bošković, Željko. 2005. Left branching extraction, structure of NP, and scrambling. In
The free word order phenomenon: Its syntactic sources and diversity, ed. J. Sabel and
M. Saito. Berlin: Mouton de Gruyter.
Bošković, Željko, and Daiko Takahashi. 1998. Scrambling and last resort. Linguistic
Inquiry 29:347–366.
Bresnan, Joan. 1972. On sentence stress and syntactic transformations. In Contributions to
Generative Phonology, ed. Michael Brame, 73–107. Oxford of Texas Press.
Büring, Daniel. 2006. Focus projection and default prominence. In The architecture of
focus, ed. Valéria Molnár and Susanne Winkler, 321–346. Berlin–New York: Mouton
De Gruyter.
Büring, Daniel, and Katharina Hartmann. 2001. The syntax and semantics of focussensitive
particles in German. Natural language and linguistic theory 19:229–281.
Cardinaletti, Anna, and Michal Starke. 1999. The typology of structural deficiency: A case
study of the three classes of pronouns. In Clitics in languages of Europe, ed. Henk C. van
Riemsdijk, George Bosong, and Bernard Comrie, 145–234. Berlin; New York: Mouton
de Gruyter.
Chafe, Wallace R. 1976. Givenness, contrastiveness, definiteness, subjects and topics. In
Subjects and topics, 25–56. New York: Academic Press.
Choi, Hye-Won. 1999. Optimizing structure in context. Stanford, California: CSLI Publications.
Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press.
Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The view from
building 20: Essays in linguistics in honor of Sylvain Bromberger, ed. Kenneth Hale and
Samuel Jay Kayser. Cambridge, MA: MIT Press.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by Step, ed. R. Martin,
D. Michaels, and J. Urigereka. Cambridge, MA: MIT Press.
143

Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A life in language, 1–52.
Cambridge, MA: MIT Press.
Chomsky, Noam. 2004a. Beyond explanatory adequacy. In Structures and Beyond, ed.
A. Belletti. New York: Oxford University Press.
Chomsky, Noam. 2004b. Class lectures; fall 2004.
Chomsky, Noam. 2005. On phases. MIT ms.
Cinque, Guglielmo. 1990. Type of A’-dependencies. Cambridge, Mass.: MIT Press.
Cinque, Guglielmo. 1993. A null theory of phrasal and compound stress. Linguistic inquiry
24:239–267.
Cinque, Guglielmo, ed. 2002. Functional structure in DP and IP. The cartography of
syntactic structures, volume 1. Oxford: Oxford University Press.
Citko, Barbara. 2006. Interaction between Across-the-Board wh-movement and leftbranching
extraction. Syntax 9:225–247.
Collins, Chris. 2002. Eliminating labels. In Derivation and explanation in the Minimalist
Program, ed. S.D. Epstein and T.D. Seely. Blackwell.
Corver, Norbert, and Henk van Riemsdijk, ed. 1994. Studies on scrambling. Movement and
non-movement approaches to free word-order phenomena. Berlin, New York: Mouton
de Gruyter.
Daneš, František. 1954. Příspěvek k rozboru významové výstavby výpovědi [a contribution
to the analysis of the semantic structure of the utterance]. Studie a práce lingvistické
I:263–274.
Daneš, František. 1957. Intonace a věta ve spisovné češtině [Intonation and sentence in
Standard Czech]. Praha: Nakladatelství Československé akademie věd.
Daneš, František. 1974. Functional sentence perspective and the organisation of the text.
In Papers on functional sentence perspective, ed. František Daneš, 106–128. Prague:
Academia.
Diesing, Molly. 1990. Verb movement and the subject position in Yiddish. Natural Language
and Linguistic Theory 8:41–79.
Diesing, Molly. 1992. Indefinites. Cambridge, Mass.: MIT Press.
den Dikken, Marcel. To Appear. Phase extension. Theoretical Linguistics .
Dokulil, Miloš, and František Daneš. 1958. K tzv. významové a mluvnické stavbě
věty [On the so called semantics and grammatical structure of the sentence]. In O
vědeckém poznání soudobých jazyků, ed. Antonín Dostál, 231–246. Praha: Nakladatelství
Československé akademie věd.
144

Dotlačil, Jakub. 2004. The syntax of infinitives in Czech. Master Thesis, Universitetet i
Tromsø.
Downing, Laura J., Luzt Marten, and Sabine Zerbian, ed. 2006. Papers in Bantu grammar
and description. ZAS papers in linguistics 43.
Elbourne, Paul D. 2005. Situations and Individuals. MIT Press.
Enç, Mürvet. 1991. The semantics of specificity. Linguistic inquiry 22:1–25.
Erteschik-Shir, Nomi. 1997. The dynamics of focus structure. CMambridge: Cambridge
University Press.
Farkas, Donka F. 2002. Specificity distinctions. Journal of Semantics 19:213–243.
Firbas, Jan. 1964. On defining the theme in functional sentence perspective. Travaux
linguistiques de Prague 1:267–280.
Firbas, Jan. 1992. Functional sentence perspective in written and spoken communication.
Cambridge, New York: Cambridge University Press.
Fox, Danny. 1995. Economy and scope. Natural language semantics 3:283–341.
Fox, Danny. 1999. Reconstruction, binding theory, and the interpretation of chains. Linguistic
Inquiry 30:157–196.
Fox, Danny. 2000. Economy and semantic interpretation. Cambridge, Mass.: MIT Press.
Fox, Danny, and David Pesetsky. 2005. Cyclic linearization of syntactic structure. Theoretical
Linguistics 31:1–45.
Franks, Steven, and Tracy Holloway King. 2000. A handbook of Slavic clitics. New York,
Oxford: Oxford University Press.
Gebauer, Jan. 1900. Příruční mluvnice jazyka českého pro učitele a studium soukromé
[Practical grammmar of Czech for teachers and private study]. Praha: Nákladem F.
Tempského.
van Geenhoven, Veerle. 1998. Semantic incorporation and indefinite descriptions: Semantic
and syntactic aspects of noun incorporation in West Greenlandic. Stanford: CSLI
Publications.
Grewendorf, Günther, and Wolfgang Sternfeld, ed. 1990. Scrambling and barriers. Amsterdam,
Philadelphia: John Benjamins Publisher.
Haider, Hubert. 1988. θ-tracking system – Evidence from German. In Configurationality,
ed. L. Maracz and P. Muysken, 185–206. Dordrecht: Foris.
Hajičová, Eva. 1973. Negation and topic vs. comment. Philologica Pragensia 16:81–93.
145

Hajičová, Eva. 1974. Negace a presupozice ve významové stavbě věty [Negation and presupposition
in the meaning structure of a sentence]. Prague: Academia.
Hajičová, Eva, Barbara H. Partee, and Petr Sgall. 1998. Topic-focus articulation, tripartite
structures, and semantic content. Dordrecht–Boston–London: Kluwer Academic
Publishers.
Hale, Ken, and Samuel Jay Keyser. 2002. Prelogomenon to a theory of argument structure.
Cambridge, Mass.: MIT Press.
Halliday, Michael A. K. 1967. Notes on transitivity and theme in English,Part II. Journal
of Linguistics 3:199–244.
Hausenblas, Karel. 1964. On characterization and classification of discourse. Travaux
Linguistiques de Prague 1:67–84.
Heck, Fabian, and Gereon Müller. 2006. Extremely local optimization. [Ms., Universität
Leipzig].
Heim, Irene. 1991. Artikel und Definitheit. In Semantik: Ein internationales Handbuch
der zeitgenössischen Forschung, 487–535. Berlin: Mouton de Gruyter.
Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford:
Blackwell.
Heycock, Caroline, and Anthony Kroch. 1993. Verb movement and coordination in a dynamic
theory of licensing. Groninger Arbeiten zur Germanistischen Linguistik 36:75–
102.
Hlavsa, Zdeněk. 1975. Denotace objektu a její prostředky v současné češtině [Denotation
of an object and its means in current Czech]. Prague: Academia.
Holmberg, Anders. 1986. Word order and syntactic features. Doctoral Dissertation, University
of Stockholm.
Horn, Laurence. 1972. On the semantic propertis of the logical operators in English. Doctoral
Dissertation, UCLA.
van der Hulst, Harry, ed. 1999. Word prosodic systems in the languages of Europe. Berlin,
New York: Mouton de Gruyter.
Iatridou, Sabine. 1995. Clitics and island effects. In Penn working papers in linguistics,
ed. Roumyana Izvorski and Victoria Tredinnick, volume 2, 11–30. University of Pennsylvania.
Iatridou, Sabine. 2002. What I’ve learned about reconstruction in A-chains [MIT ms.].
Johnson, Kyle. 2002. Restoring exotic coordinations to normalcy. Linguistic inquiry 33:97–
156.
146

Junghanns, Uwe. 1999. Generative Beschreibung periphrasticher Konstruktionen des
Tschechischen. In Linguistische Beiträge zur Slavistik aus Deutschland and Österreich.
VII. JungslavistInnen-Treffen, Tübingen / Blaubeuren 1998., ed. T. Anstatt, R. Meyer,
and E. Seitz. München: Saguer.
Junghanns, Uwe, and Gerhild Zybatow. 1997. Syntax and information structure of Russian
clauses. In Annual workshop on formal approaches to Slavic linguistics. The Cornell
meeting 1995 (= Michigan Slavic Materials; 39), ed. E. W. Browne et al. Ann Arbor:
Browne, E.W. et al.
Karimi, Simin, ed. 2003. Word order and scrambling. Blackwell Publishing.
Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.
Kiss, Katalin É. 1994. Scrambling as the base-generation of random complement order.
In Studies on scrambling, ed. Norbert Corver and Henk van Riemsdjijk, 61–100. Berlin,
New York: Mouton de Gruyter.
Ko, Heejeong. 2007. Asymmetries in scrambling and cyclic linearization. Linguistic Inquiry
38:49–83.
Krifka, Manfred. 2001. Non-novel indefinites in adverbial quantification. In Logical perspectives
on language and information, ed. Cleo Condoravdi and Gerard Renardel der
Lavalette, 1–40. Stanford: CSLI Press.
Krifka, Manfred. 2006a. Association with focus phrases. In The architecture of focus, ed.
Valéria Molnár and Susanne Winkler, 105–136. Berlin–New York: Mouton de Gruyter.
Krifka, Manfred. 2006b. Can focus accenting be eliminated in favor of deaccenting Given
constituents. In Proceedings of the Ninth Symposium on logic and language, Hungary,
August 23–26.
Kruijff-Korbayová, Ivana, and Mark Steedman. 2003. Discourse and information structure.
Journal of Logic, Language and Information 12:249–259.
Kusumoto, Kiyomi. 2005. On the quantification over times in natural language. Natural
language semantics 13:317–357.
Kučerová, Ivona. 2005. The T-Extension condition. In Procedings of the 24th West Coast
Conference on Formal Linguistics, ed. John Alderete et al., 227–235. Somerville, MA:
Cascadilla Proceedings Project.
Ladd, D. Robert. 1980. The structure of intonational meaning. Cambridge: Cambridge
University Press.
Ladd, D. Robert. 1996. Intonational Phonology. Cambridge: Cambridge University Press.
Larson, Richard. 1988. On the double object construction. Linguistic Inquiry 19:335–391.
147

Lasnik, Howard. 1999. Chains of arguments. In Working Minimalism, ed. David Epstein
and Norbert Hornstein. Cambridge: MIT Press.
Lasnik, Howard, and Mamoru Saito. 1992. Move α. Conditions on its application and
output. Cambridge, MA: MIT Press.
Lavine, James, and Robert Freidin. 2002. The subject of defective T(ense) in Slavic. Journal
of Slavic linguistics 10:251–287.
Lebeaux, David. 1988. Language acquisition and the form of grammar. Doctoral Dissertation,
University of Massachusetts at Amherst.
Lebeaux, David. 1998. Where does binding theory apply? Technical report, NEC.
Magri, Giorgio. 2007. A theory of individual level predicates based on blind scalar implicatures
[MIT ms.].
Mahajan, Anoop Kumar. 1990. The A/A-bar distinction and movement theory. Doctoral
Dissertation, MIT.
Mathesius, Vilém. [1929] 1983. Functional linguistics. In Praguiana: Some basic and wellknown
aspects of the Prague Linguistics School, ed. J. Vachek, 121–142. Amsterdam:
John Benjamins.
Mathesius, Vilém. 1931. K dynamické linii české věty [To the dynamic line of a Czech
sentence]. Časopis pro moderní filologii 17:71ff.
Mathesius, Vilém. 1937. Mluvní takt a některé problémy příbuzné [Prosodic tact and some
related problems]. Slovo a slovesnost 3:193–199.
Mathesius, Vilém. 1939. O tak zvaném aktuálním členění věty [On the so called
topic/comment articulation of the sentence]. Slovo a slovesnost 5:171–174.
Meyer, Roland. 2003. On multiple wh-fronting and wh-clustering in Czech. In Formal
approaches to Slavic linguistics XII: The Amherst meeting, ed. W. Browne, E. Kim, and
B. Partee, 393–412. Ann Arbor: Michigan Slavic Publications.
Meyer, Roland. 2004. Syntax der Ergänzungsfrage. Empirische Untersuchungen am Russischen,
Polnischen und Tschechischen [syntax of wh-questions. Empirical studies in
russian, polish and czech]. München: Verlag Otto Sagner.
Miyagawa, Shigeru. 1997. Against optional scrambling. Linguistic Inquiry 28:1–25.
Neeleman, Ad. 1994. Scrambling as a D-structure phenomenon. In Studies on scrambling,
ed. Norbert Corver and Henk van Riemsdjijk, 387–429. Berlin, New York: Mouton de
Gruyter.
Neeleman, Ad, and Hans van de Koot. 2006. The nature of discourse templates. [UCL
ms.].
148

Neeleman, Ad, and Tanya Reinhart. 1998. Scrambling and the PF interface. In The projection
of arguments, ed. M. Butt and W. Geuder, 309–353. CSLI.
Nevins, Andrew, and Pranav Anand. 2003. Some AGREEment matters. In Proceedings of
WCCFL 22, ed. G. Garding and M. Tsujimura, 370–383. Somerville, Mass.: Cascadilla
Press.
Nissenbaum, Jon. 1998. Movement and derived predicates: Evidence from Parasitic Gaps.
In The interpretive tract, ed. Uli Sauerland and Orin Percus, 247–295. Cambridge, MA:
MITWPL.
Nissenbaum, Jonathan W. 2000. Investigations of covert phrase movement. Doctoral Dissertation,
MIT.
Novák, Pavel. 1959. O prostředcích aktuálního členění [On the means of topic-focus articulation].
AUC Philologica 1:9–15.
Ogihara, Toshiyuki. 1996. Tense, Attitudes, and Scope. Springer.
Palková, Zdena. 1994. Fonetika a fonologie češtiny [Phonetics and phonology of Czech].
Praha: Karolinum.
Partee, Barbara H., and Mats Rooth. 1983. Generalized conjunction and type ambiquity. In
Meaning, use and the interpretation of language, ed. Rainer Bäurle, Christoph Schwarze,
and Arnim von Stechow, 361–393. Walter de Gruyter & Co.
Percus, Orin. 2006. Anti-presuppositions. In Theoretical and empirical studies of reference
and anaphora: Toward the establishment of generative grammar as an empirical
science, ed. A. Ueyamada, 52–73. Japan Society for the Promotion of Science.
Pesetsky, David. 1982. Paths and categories. Doctoral Dissertation, MIT.
Pesetsky, David, and Esther Torrego. 2006. Probes, goals and syntactic categories. In
Proceedings of the Seventh Tokyo conference on psycholinguistics, ed. Yukio Otsu, 25–
60. Tokyo: Hituzi Svobo Publishing Company.
Petr, Jan, Miloš Dokulil, Karel Horálek, Jiřina Hůrková, and Miloslava Knappová, ed.
1986. Mluvnice češtiny [Grammar of Czech]. Praha: Academia.
Petřík, Stanislav. 1938. O hudební stránce stědočeské věty [About musical properties of a
Central Bohemian sentence]. Praha: Filosofická fakulta University Karlovy.
Postal, Paul. 1966. On so-called “pronouns” in English. In Report on the Seventeenth Annual
Round Table Meeting on Linguistics and Language Studies, 177–206. Washington,
DC: Georgetown University Press.
Pylkkänen, Liina. 2002. Introducing arguments. Doctoral Dissertation, MIT, Cambridge,
MA.
Reinhart, Tanya. 1995. Focus – the PF interface [University of Utrecht, ms.].
149

Reinhart, Tanya. 1997. Quantifier scope: How labor is divided between QR and choice
functions. Linguistics and philosophy 20:335–397.
Reinhart, Tanya. 2006. Interface strategies. Optimal and costly computations. Cambridge,
MA: MIT Press.
Rezac, Milan. 2005. The syntax of clitic climbing in Czech. In Clitic and affix combinations:
Theoretical perspectives, ed. Lorie Heggie and Francisco Ordó nez. John
Benjamins Publishing Company.
Richards, Norvin. 1997. What moves where when and in which language? Doctoral
Dissertation, MIT.
Rizzi, Luigi. 1990. Relativized minimality. Cambridge, Mass.: MIT Press.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar:
Handbook of generative syntax, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer.
Rizzi, Luigi, ed. 2004. The structure of CP and IP. The cartography of syntactic structures,
volume 2. Oxford: Oxford University Press.
Romportl, Milan. 1973. Základy fonetiky [Basics of phonetics]. Praha.
Ross, J. R. 1967. Constraints on variables in syntax. Doctoral Dissertation, MIT, Cambridge,
MA.
Sabel, Joachim, and Mamoru Saito, ed. 2005. The free word order phenomenon: its syntactic
sources and diversity. Berlin, New York: Mouton de Gruyter.
Saito, Mamoru. 1989. Scrambling as semantically vacuous A-movement. In Alternative
conceptions of phrase structure, ed. Mark Baltin and Anthony Kroch, 182–200. Chicago:
University of Chicago Press.
Saito, Mamoru. 1992. Long distance scrambling in Japanese. Journal of East Asian linguistics
1:69–119.
Sauerland, Uli. 1999. Erasibility and interpretation. Syntax 2:161–188.
Sauerland, Uli. 2003. A new semantics for number. In Proceedings of SALT 13, ed.
R. Young and Y. Zhou. Cornell University, Ithaca, N.Y: CLC Publications.
Sauerland, Uli. 2005. Don’t interpret focus: Why a presuppositional account of focus fails,
and how a presuppositional account of givenness works. In Proceedings of Sinn und
Bedeutung 9, 370–384. University of Nijmegen, Netherlands.
Schlenker, Philippe. 2006. ‘maximize Presupposition’ and Gricean reasoning [UCLA &
Institut Jean-Nicod ms.].
Schwarzschild, Roger. 1999. GIVENness, AvoidF and other constraints on the placement
of accent. Natural language semantics 7:141–177.
150

Selkirk, Elizabeth O. 1995. Sentence prosody: Intonation, stress, and phrasing. In The
handbook of phonology, ed. J.A.Goldsmith, 550–569. London: Blackwell.
Sgall, Petr. 1967. Functional sentence perspective in a generative description. Prague
Studies in Mathematical Linguistics 2:203–225.
Sgall, Petr, Eva Hajičová, and Eva Buráňová. 1980. Aktuální členění věty v češtině
[Topic/focus articulation of the Czech sentences]. Praha: Academia.
Sgall, Petr, Eva Hajičová, and Jarmila Panevová. 1986. The meaning of the sentence in
its semantic and pragmatic aspects. Dordrecht, Boston, Lancaster, Tokyo: D. Reidel
Publishing Company.
Sigurðsson, Halldór Ármann. 2004. Agree and agreement: Evidence from Germanic. Focus
on Germanic Typology, Studia Typologica 6:61–103.
Skaličková, Alena. 1956. K otázce větného přízvuku v češtině [To the question of sentential
stress in Czech]. AUC Philologica 55–86.
Stalnaker, Robert. 1973. Presuppositions. Journal of Philosophical Logic 2:447–457.
Stalnaker, Robert. 1974. Pragmatic presupposition. In Semantics and philosophy, ed.
Milton Munitz and Peter Unger, 197–213. New York: New-York University Press.
Stalnaker, Robert C. 2002. Common ground. Linguistics & Philosophy 25:701–721.
Sturgeon, Anne. 2006. The syntax and pragmatics of contrastive topic in Czech. Doctoral
Dissertation, UC Santa Cruz.
Svenonius, Peter. 1994. C-selection as feature-checking. Studia linguistica 48:133–155.
Svoboda, Aleš. 1984. České slovosledné pozice z pohledu aktuálního členění [Czech word
order positions from the point of view of topic focus articulation]. Slovo a slovesnost
45:22–34, 88–103.
Szendrői, Kriszta. 2003. A stress-based approach to the syntax of Hungarian focus. The
linguistic review 20:37–78.
Tateishi, Koichi. 1994. The syntax of ‘subjects’. Stanford: CSLI.
Trávníček, František. 1937. Základy československého slovosledu [Basics of Czechoslovak
word order]. Slovo a slovesnost 3:78–86.
Trávníček, František. 1939. Slovosled při důrazu [Word order under emphasis]. Slovo a
slovesnost 5:131–144.
Truckenbrodt, Hubert. 1995. Phonological phrases: Their relation to syntax, focus, and
prominence. Doctoral Dissertation, MIT.
Vallduví, Enric. 1990. The intonation component. Doctoral Dissertation, University of
Pennsylvania.
151

Vallduví, Enric. 1992. The informational component. New York, London: Garland Publishing,
Inc.
Veselovská, Ludmila. 1995. Phrasal Movement and X-Morphology: Word Order Parallels
in Czech and English Nominal and Verbal Projections. Doctoral Dissertation, Palacký
University, Olomouc.
Veselovská, Ludmila. 1998. Possessive movement in the Czech nominal phrase. Journal
of Slavic linguistics 6:255–300.
Veselovská, Ludmila. 2004. The extended verbal projection in Czech: Three variants of
the verb ‘be’. Presented at FDSL 5. University of Potsdam, Germany.
Veselovská, Ludmila, and Petr Karlík. 2004. Analytic passives in Czech. Zeitschrift für
Slavistik 2:163–243.
Vos, Riet, and Ludmila Veselovská. 1999. Clitic questionnaire. In Clitics in the languages
of Europe, ed. Henk van Riemsdijk, 891–1009. Berlin, New York: Mouton de Gruyter.
Šmilauer, Vladimír. 1960. Pořádek slov [Word order]. In O češtině pro Čechy, ed.
František Daneš et al., 247–270. Praha: Novinářský studijní ústav.
Wagner, Michael. 2005. Prosody and recursion. Doctoral Dissertation, MIT.
Wagner, Michael. To appear. NPI-licensing and focus movement. In Proceedings of SALT
XV. Cornell University., ed. Effi Georgala and Jonathan Howell. Cornell University.
Wagner, Michael. To appear a. Association by movement: Evidence from NPI-licensing.
Natural language semantics .
Wagner, Michael. To appear b. Givenness and locality. In Proceedings of SALT XVI. Ithaca,
NY: CLC Publications.
Webelhuth, Gert. 1989. Syntactic saturation phenomena and the modern Germanic languages.
Doctoral Dissertation, University of Massachusetts, Amherst.
Williams, Edwin. 1984. There-insertion. Linguistic Inquiry 15:131–153.
Williams, Edwin. 1997. Blocking and anaphora. Linguistic Inquiry 28:577–628.
Wurmbrand, Susi. 2006. Tense topics: The future of infinitives [Colloquium talk, MIT,
December 2006].
Wurmbrand, Susi. To appear. Infinitives are tenseless. Proceedings of the 30th Annual
Penn linguistics colloquium. Penn Working Papers in Linguistics 13.
Zerbian, Sabine. 2006. Expression of information structure in the Bantu language Northern
Sotho. ZAS Papers in Linguistics 45.
152

Zikánová, Šárka. 2006. Slovosled ve starší češtině (1500-1620): Postavení syntetického
přísudku ve větě hlavní [Word order in older Czech (1500-1620): The position of the
synthetic predicate in matrix clauses]. Doctoral Dissertation, Charles University, Prague.
Zubizarreta, Maria Luisa. 1998. Prosody, focus, and word order. Cambridge, Mass.: MIT
Press.
153